DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Information Disclosure Statement
The information disclosure statement (IDS) submitted on July 5, 2023 is in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statement is being considered by the examiner.
Claim Objections
Claim 5 and 16 are objected to because of the following informalities: In claim 5 and 16, line 2, " performing a check to determine is there exists an existing item", should read "performing a check to determine if there exists an existing item". In claim 5 and 16, line 3, " similarity(intent_embeddingsi, item_embeddingj)", should read "a similarity between an intent embedding i and an item embedding j". Appropriate correction is required.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 1-20 rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea (mental process) without significantly more.
Claim 1:
Regarding claim 1, in step 1 of the 101-analysis set forth in MPEP 2106, the claim recites “A method comprising: encoding, using at least one hardware processor…”, and a method is a process which is one of the four statutory categories of invention.
In step 2A prong 1 of the 101-analysis set forth in the MPEP 2106, the examiner has determined that the following limitations recite a process that, under the broadest reasonable interpretation, covers a mental process but for recitation of generic computer components:
“encoding…, a first intent from customer provided data as an intent embedding” (this is a mental process, a person can mentally evaluate an intent from provided list of data and write a code as a vector representation considered as an intent embedding, see MPEP § 2106.04(a)(2)(III)),
“comparing…, the intent embedding of the first intent and an intent embedding corresponding to one or more items of a training workspace to generate a similarity score” (this is a mental process, a person can mentally evaluate the vector representation for the intent from the data to other vector representations and do a mathematical formula to compute a distance/score of how similar the vectors representation of the first intent is to other vector representations of a training dataset by use of pen and paper, see MPEP § 2106.04(a)(2)(III)),
“mapping…, the first intent to a similar item of the one or more items and incrementing a corresponding count of the similar item by one in response to the similarity score being greater than a given threshold” (this is a mental process, a person can mentally evaluate matching/mapping the vector representation for the intent from the data to other vector representations which can be linked to an item and incrementing the count of this item based on the chosen mathematical formula to compute a score being greater than a chosen threshold, see MPEP § 2106.04(a)(2)(III)),
“creating…, a matrix based on the similarity score, the created matrix including selected training workspaces” (this is a mental process, a person can mentally evaluate creating a matrix which can include the training dataset mentioned and making this matrix based on the chosen mathematical formula to compute a score, see MPEP § 2106.04(a)(2)(III)),
If claim limitations, under the broadest reasonable interpretation, covers performance of the limitations as a mental process but for the recitation of generic computer components, then it falls within the mental process grouping of abstract ideas. Accordingly, the claim “recites” an abstract idea.
In step 2A prong 2 of the 101-analysios set forth in MPEP 2106, the examiner has determined that the following additional elements do not integrate this judicial exception into a practical application:
“using at least one hardware processor” (Using processors is considered generic computer component being used as tool to perform functions of the judicial exception – see MPEP § 2106.05(f)),
“training, using the at least one hardware processor, at least a first machine learning model using one or more of the selected training workspaces of the created matrix” (training a machine learning model is considered mere instructions to apply an exception using generic computer – see MPEP § 2106.05(f)),
“training, using the at least one hardware processor, at least a second machine learning model using the created matrix” (As previously mentioned, training a machine learning model is considered mere instructions to apply an exception using generic computer – see MPEP § 2106.05(f)),
“facilitating, using the at least one hardware processor, deployment of the at least second machine learning model for performing inferencing.” (facilitating deployment of a machine learning model is considered mere instructions to apply an exception using generic computer – see MPEP § 2106.05(f))
Since the claim as a whole, looking at the additional elements individually and in combination, does not contain any other additional elements that are indicative of integration into a practical application, the claim is “directed” to an abstract idea.
In step 2B of the 101-analysis set forth in the 2019 PEG, the examiner has determined that the claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception.
As discussed above, additional element v recites a generic computer component being used as tool to perform functions of the judicial exception, and additional elements vi, vii, and viii recite mere instructions to apply the judicial exception using generic computer components, which are not indicative of significantly more.
Considering the additional elements individually and in combination, and the claim as a whole, the additional elements do not provide significantly more than the abstract idea. Therefore, the claim is not patent eligible.
Claim 2:
Regarding claim 2, it is dependent upon claim 1, and thereby incorporates the limitations of, and corresponding analysis applied to claim 1. Further, claim 2 recites the following additional element:
“The method of claim 1, further comprising performing inferencing using the deployed at least second trained machine learning model” (In step 2A, prong 2, this is considered mere instructions to apply an exception using generic computer with the generate task-specific output operation performed by any generic computer, see MPEP § 2106.05(f)). (In step 2B, this is also considered mere instructions to apply an exception using generic computer - see MPEP § 2106.05(f)).
Since the claim does not recite additional elements that either integrate the judicial exception into a practical application, nor provide significantly more than the judicial exception, the claim is not patent eligible.
Claim 3:Regarding claim 3, it is dependent upon claim 1, and thereby incorporates the limitations of, and corresponding analysis applied to claim 1. Further claim 3 recites the following abstract idea:
“method of claim 1, further comprising repeating the comparing operation and creating a new item corresponding to a second intent in response to the similarity score being less than the given threshold and setting a count corresponding to the new item to one” (this is considered a mental process, since a person can mentally evaluate and repeat the comparing limitation from claim 1 to identify a new item based on the chosen mathematical formula to compute a score being less than the threshold of an already established item and incrementing the count for this new item by use of pen and paper, see MPEP § 2106.04(a)(2)(III)),
If claim limitations, under their broadest reasonable interpretation, covers performance of the limitations as a mental process but for the recitation of generic compute components, then it falls within the mental process grouping of abstract ideas. Accordingly, the claim “recites” an abstract idea.
Since the claim does not recite additional elements that either integrate the judicial exception into a practical application, nor provide significantly more than the judicial exception, the claim is not patent eligible.
Claim 4:
Regarding claim 4, it is dependent upon claim 3, and thereby incorporates the limitations of, and corresponding analysis applied to claim 3. Further claim 4 recites the following abstract ideas:
“The method of claim 3, further comprising: clustering client log data into candidate intents and mapping the candidate intents to the items of the matrix” (this is considered a mental process, since a person can mentally evaluate clustering client log data to candidate intents and mapping/matching these intents to items of the matrix established in claim 1, see MPEP § 2106.04(a)(2)(III)),
“creating at least one intent recommendation based on intents corresponding to the items of the training workspace” (this is considered a mental process, since a person can mentally evaluate identifying and creating a recommendation based on intents in correspondence to items from training dataset, see MPEP § 2106.04(a)(2)(III)),
If claim limitations, under their broadest reasonable interpretation, covers performance of the limitations as a mental process but for the recitation of generic compute components, then it falls within the mental process grouping of abstract ideas. Accordingly, the claim “recites” an abstract idea.
Since the claim does not recite additional elements that either integrate the judicial exception into a practical application, nor provide significantly more than the judicial exception, the claim is not patent eligible.
Claim 5:
Regarding claim 5, it is dependent upon claim 3, and thereby incorporates the limitations of, and corresponding analysis applied to claim 3. Further claim 5 recites the following abstract idea:
“The method of claim 3, wherein the comparing operation further comprises performing a check to determine is there exists an existing item with a similarity(intent_embeddingsi, item_embeddingj) that is greater than the given threshold” (this is considered a mental process, since a person can mentally evaluate using the comparing limitation from claim 1 and mentally checking if there is already an existing item with similarity that is greater than the chosen threshold, see MPEP § 2106.04(a)(2)(III)),
If claim limitations, under their broadest reasonable interpretation, covers performance of the limitations as a mental process but for the recitation of generic compute components, then it falls within the mental process grouping of abstract ideas. Accordingly, the claim “recites” an abstract idea.
Since the claim does not recite additional elements that either integrate the judicial exception into a practical application, nor provide significantly more than the judicial exception, the claim is not patent eligible.
Claim 6:
Regarding claim 6, it is dependent upon claim 3, and thereby incorporates the limitations of, and corresponding analysis applied to claim 3. Further claim 6 recites the following abstract idea:
“The method of claim 3, further comprising repeating the encoding, comparing, mapping, creating the new item, and creating the matrix operations, for each workspace to update the matrix with each pair of training workspace and item” (this is considered a mental process, since a person can mentally evaluate using the encoding, comparing, mapping, creation of a new item, and creation of the matrix limitations from claim 1, to then make updates to the created matrix with each training dataset and item, see MPEP § 2106.04(a)(2)(III)),
If claim limitations, under their broadest reasonable interpretation, covers performance of the limitations as a mental process but for the recitation of generic compute components, then it falls within the mental process grouping of abstract ideas. Accordingly, the claim “recites” an abstract idea.
Since the claim does not recite additional elements that either integrate the judicial exception into a practical application, nor provide significantly more than the judicial exception, the claim is not patent eligible.
Claim 7:
Regarding claim 7, it is dependent upon claim 3, and thereby incorporates the limitations of, and corresponding analysis applied to claim 3. Further claim 7 recites the following abstract ideas:
“The method of claim 3, further comprising: mapping a given client workspace into the matrix” (this is considered a mental process, since a person can mentally evaluate mapping/matching a client dataset into a matrix, see MPEP § 2106.04(a)(2)(III)),
“grouping utterances of client log data into clusters and mapping the clusters into items of the similar training workspaces using the corresponding embeddings” (this is considered a mental process, since a person can mentally evaluate grouping utterances of client log data into clusters/groupings and mapping/matching these groupings/clusters into items of the similar training datasets using the corresponding vector representations considered as embeddings, see MPEP § 2106.04(a)(2)(III)),
“recommending the first and second intents corresponding to the mapped items that exist in the similar training workspace and are absent from the client workspace” (this is considered a mental process, since a person can mentally evaluate and recommend a first and second intent in correspondence to mapped/matched items that exist from the similar training dataset and that are missing from the client dataset, see MPEP § 2106.04(a)(2)(III)),
If claim limitations, under their broadest reasonable interpretation, covers performance of the limitations as a mental process but for the recitation of generic compute components, then it falls within the mental process grouping of abstract ideas. Accordingly, the claim “recites” an abstract idea.
Further, claim 7 recites the following additional element:
“conducting a search of the training workspace that contains a similar set of items compared to the given client workspace” (In step 2A, prong 2, this is considered mere instructions to apply an exception using generic computer with the generate task-specific output operation performed by any generic computer, see MPEP § 2106.05(f)). (In step 2B, this is also considered mere instructions to apply an exception using generic computer - see MPEP § 2106.05(f)).
Since the claim does not recite additional elements that either integrate the judicial exception into a practical application, nor provide significantly more than the judicial exception, the claim is not patent eligible.
Claim 8:
Regarding claim 8, it is dependent upon claim 3, and thereby incorporates the limitations of, and corresponding analysis applied to claim 3. Further claim 8 recites the following abstract ideas:
“The method of claim 3, further comprising: mapping a given client workspace into the matrix” (this is considered a mental process, since a person can mentally evaluate mapping/matching a client dataset into a matrix, see MPEP § 2106.04(a)(2)(III)),
“recommending the first and second intents corresponding to the mapped items that exist in the similar training workspace and are absent from the client workspace” (this is considered a mental process, since a person can mentally evaluate recommending first and second intents in correspondence to mapped/matched items that exist from the similar training dataset and that are missing from the client dataset, see MPEP § 2106.04(a)(2)(III)),
If claim limitations, under their broadest reasonable interpretation, covers performance of the limitations as a mental process but for the recitation of generic compute components, then it falls within the mental process grouping of abstract ideas. Accordingly, the claim “recites” an abstract idea.
Further, claim 8 recites the following additional element:
“conducting a search of the training workspace(s) that contain a similar set of items in the matrix compared to the given client workspace” (In step 2A, prong 2, this is considered mere instructions to apply an exception using generic computer with the generate task-specific output operation performed by any generic computer, see MPEP § 2106.05(f)). (In step 2B, this is also considered mere instructions to apply an exception using generic computer - see MPEP § 2106.05(f)).
Since the claim does not recite additional elements that either integrate the judicial exception into a practical application, nor provide significantly more than the judicial exception, the claim is not patent eligible.
Claim 9:
Regarding claim 9, it is dependent upon claim 3, and thereby incorporates the limitations of, and corresponding analysis applied to claim 3. Further claim 9 recites the following abstract ideas:
“The method of claim 3, further comprising: grouping utterances of client log data into clusters and mapping the clusters into items of the training workspace using the corresponding embeddings” (this is considered a mental process, since a person can mentally evaluate grouping utterances of client log data into clusters/groupings and mapping/matching these groupings/clusters into items of the training datasets using the corresponding vector representations considered as embeddings, see MPEP § 2106.04(a)(2)(III)),
“recommending the first and second intents corresponding to the mapped items that exist in the training workspace” (this is considered a mental process, since a person can mentally evaluate recommending first and second intents in correspondence to mapped/matched items that exist from the training dataset, see MPEP § 2106.04(a)(2)(III)),
If claim limitations, under their broadest reasonable interpretation, covers performance of the limitations as a mental process but for the recitation of generic compute components, then it falls within the mental process grouping of abstract ideas. Accordingly, the claim “recites” an abstract idea.
Since the claim does not recite additional elements that either integrate the judicial exception into a practical application, nor provide significantly more than the judicial exception, the claim is not patent eligible.
Claim 10:Regarding claim 10, it is dependent upon claim 2, and thereby incorporates the limitations of, and corresponding analysis applied to claim 2. Further claim 10 recites the following abstract idea:
“The method of claim 2, wherein the inferencing is performed to generate a recommendation for a conversational artificial intelligence task” (this is considered a mental process, since a person can mentally evaluate the inferencing of claim 2 for creating a recommendation intended for a conversational artificial intelligence task, see MPEP § 2106.04(a)(2)(III)),
If claim limitations, under their broadest reasonable interpretation, covers performance of the limitations as a mental process but for the recitation of generic compute components, then it falls within the mental process grouping of abstract ideas. Accordingly, the claim “recites” an abstract idea.
Since the claim does not recite additional elements that either integrate the judicial exception into a practical application, nor provide significantly more than the judicial exception, the claim is not patent eligible.
Claim 11:
Regarding claim 11, in step 1 of the 101-analysis set forth in MPEP 2106, the claim recites “A computer program product, comprising: one or more tangible computer-readable storage media and program instructions stored on at least one of the one or more tangible computer-readable storage media, the program instructions executable by a processor, the program instructions…”, and a computer program product is a machine which is one of the four statutory categories of invention.
In step 2A prong 1 of the 101-analysis set forth in the MPEP 2106, the examiner has determined that the following limitations recite a process that, under the broadest reasonable interpretation, covers a mental process but for recitation of generic computer components:
“encoding, a first intent from customer provided data as an intent embedding” (this is a mental process, a person can mentally evaluate an intent from provided list of data and write a code as a vector representation considered as an intent embedding, see MPEP § 2106.04(a)(2)(III)),
“comparing, the intent embedding of the first intent and an intent embedding corresponding to one or more items of a training workspace to generate a similarity score” (this is a mental process, a person can mentally evaluate the vector representation for the intent from the data to other vector representations and do a mathematical formula to compute a distance/score of how similar the vectors representation of the first intent is to other vector representations of a training dataset by use of pen and paper, see MPEP § 2106.04(a)(2)(III)),
“mapping, the first intent to a similar item of the one or more items and incrementing a corresponding count of the similar item by one in response to the similarity score being greater than a given threshold” (this is a mental process, a person can mentally evaluate matching/mapping the vector representation for the intent from the data to other vector representations which can be linked to an item and incrementing the count of this item based on the chosen mathematical formula to compute a score being greater than a chosen threshold, see MPEP § 2106.04(a)(2)(III)),
“creating, a matrix based on the similarity score, the created matrix including selected training workspaces” (this is a mental process, a person can mentally evaluate creating a matrix which can include the training dataset mentioned and making this matrix based on the chosen mathematical formula to compute a score, see MPEP § 2106.04(a)(2)(III)),
If claim limitations, under the broadest reasonable interpretation, covers performance of the limitations as a mental process but for the recitation of generic computer components, then it falls within the mental process grouping of abstract ideas. Accordingly, the claim “recites” an abstract idea.
In step 2A prong 2 of the 101-analysios set forth in MPEP 2106, the examiner has determined that the following additional elements do not integrate this judicial exception into a practical application:
“A computer program product, comprising: one or more tangible computer-readable storage media and program instructions stored on at least one of the one or more tangible computer-readable storage media, the program instructions executable by a processor” (Using processors and a tangible computer-readable storage media with program instructions stored on it is considered mere instructions to apply an exception using generic computer – see MPEP § 2106.05(f)),
“training at least a first machine learning model using one or more of the selected training workspaces of the created matrix” (training a machine learning model is considered mere instructions to apply an exception using generic computer – see MPEP § 2106.05(f)),
“training, using the at least one hardware processor, at least a second machine learning model using the created matrix” (As previously mentioned, training a machine learning model and using a processor is considered mere instructions to apply an exception using generic computer – see MPEP § 2106.05(f)),
“facilitating deployment of the at least second machine learning model for performing inferencing.” (facilitating deployment of a machine learning model is considered mere instructions to apply an exception using generic computer – see MPEP § 2106.05(f))
Since the claim as a whole, looking at the additional elements individually and in combination, does not contain any other additional elements that are indicative of integration into a practical application, the claim is “directed” to an abstract idea.
In step 2B of the 101-analysis set forth in the 2019 PEG, the examiner has determined that the claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception.
As discussed above, additional elements v, vi, vii, and viii recite mere instructions to apply the judicial exception using generic computer components, which are not indicative of significantly more.
Considering the additional elements individually and in combination, and the claim as a whole, the additional elements do not provide significantly more than the abstract idea. Therefore, the claim is not patent eligible.
Claim 12:
Regarding claim 12, in step 1 of the 101-analysis set forth in MPEP 2106, the claim recites “A system comprising: a memory; and at least one processor, coupled to said memory, and operative to perform operations…”, and a system in this instance is a machine which is one of the four statutory categories of invention.
In step 2A prong 1 of the 101-analysis set forth in the MPEP 2106, the examiner has determined that the following limitations recite a process that, under the broadest reasonable interpretation, covers a mental process but for recitation of generic computer components:
“encoding, a first intent from customer provided data as an intent embedding” (this is a mental process, a person can mentally evaluate an intent from provided list of data and write a code as a vector representation considered as an intent embedding, see MPEP § 2106.04(a)(2)(III)),
“comparing, the intent embedding of the first intent and an intent embedding corresponding to one or more items of a training workspace to generate a similarity score” (this is a mental process, a person can mentally evaluate the vector representation for the intent from the data to other vector representations and do a mathematical formula to compute a distance/score of how similar the vectors representation of the first intent is to other vector representations of a training dataset by use of pen and paper, see MPEP § 2106.04(a)(2)(III)),
“mapping, the first intent to a similar item of the one or more items and incrementing a corresponding count of the similar item by one in response to the similarity score being greater than a given threshold” (this is a mental process, a person can mentally evaluate matching/mapping the vector representation for the intent from the data to other vector representations which can be linked to an item and incrementing the count of this item based on the chosen mathematical formula to compute a score being greater than a chosen threshold, see MPEP § 2106.04(a)(2)(III)),
“creating, a matrix based on the similarity score, the created matrix including selected training workspaces” (this is a mental process, a person can mentally evaluate creating a matrix which can include the training dataset mentioned and making this matrix based on the chosen mathematical formula to compute a score, see MPEP § 2106.04(a)(2)(III)),
If claim limitations, under the broadest reasonable interpretation, covers performance of the limitations as a mental process but for the recitation of generic computer components, then it falls within the mental process grouping of abstract ideas. Accordingly, the claim “recites” an abstract idea.
In step 2A prong 2 of the 101-analysios set forth in MPEP 2106, the examiner has determined that the following additional elements do not integrate this judicial exception into a practical application:
“A system comprising: a memory; and at least one processor, coupled to said memory, and operative to perform operations” (Using a processor and memory is considered a generic computer component being used as tool to perform functions of the judicial exception – see MPEP § 2106.05(f)),
“training at least a first machine learning model using one or more of the selected training workspaces of the created matrix” (training a machine learning model is considered mere instructions to apply an exception using generic computer – see MPEP § 2106.05(f)),
“training, at least a second machine learning model using the created matrix” (As previously mentioned, training a machine learning model is considered mere instructions to apply an exception using generic computer – see MPEP § 2106.05(f)),
“facilitating deployment of the at least second machine learning model for performing inferencing.” (facilitating deployment of a machine learning model is considered mere instructions to apply an exception using generic computer – see MPEP § 2106.05(f))
Since the claim as a whole, looking at the additional elements individually and in combination, does not contain any other additional elements that are indicative of integration into a practical application, the claim is “directed” to an abstract idea.
In step 2B of the 101-analysis set forth in the 2019 PEG, the examiner has determined that the claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception.
As discussed above, additional element v recites a generic computer component being used as tool to perform functions of the judicial exception, and additional elements vi, vii, and viii recite mere instructions to apply the judicial exception using generic computer components, which are not indicative of significantly more.
Considering the additional elements individually and in combination, and the claim as a whole, the additional elements do not provide significantly more than the abstract idea. Therefore, the claim is not patent eligible.
Claim 13:
Regarding claim 13, it is dependent upon claim 12, and thereby incorporates the limitations of, and corresponding analysis applied to claim 12. Further, claim 13 recites the following additional element:
“The system of claim 12, the operations further comprising performing inferencing using the deployed at least second trained machine learning model” (In step 2A, prong 2, this is considered mere instructions to apply an exception using generic computer with the generate task-specific output operation performed by any generic computer, see MPEP § 2106.05(f)). (In step 2B, this is also considered mere instructions to apply an exception using generic computer - see MPEP § 2106.05(f)).
Since the claim does not recite additional elements that either integrate the judicial exception into a practical application, nor provide significantly more than the judicial exception, the claim is not patent eligible.
Claim 14:Regarding claim 14, it is dependent upon claim 12, and thereby incorporates the limitations of, and corresponding analysis applied to claim 12. Further claim 14 recites the following abstract idea:
“The system of claim 12, the operations further comprising repeating the comparing operation and creating a new item corresponding to a second intent in response to the similarity score being less than the given threshold and setting a count corresponding to the new item to one” (this is considered a mental process, since a person can mentally evaluate and repeat the comparing limitation from claim 12 to identify a new item based on the chosen mathematical formula to compute a score being less than the threshold of an already established item and incrementing the count for this new item by use of pen and paper, see MPEP § 2106.04(a)(2)(III)),
If claim limitations, under their broadest reasonable interpretation, covers performance of the limitations as a mental process but for the recitation of generic compute components, then it falls within the mental process grouping of abstract ideas. Accordingly, the claim “recites” an abstract idea.
Since the claim does not recite additional elements that either integrate the judicial exception into a practical application, nor provide significantly more than the judicial exception, the claim is not patent eligible.
Claim 15:
Regarding claim 15, it is dependent upon claim 14, and thereby incorporates the limitations of, and corresponding analysis applied to claim 14. Further claim 15 recites the following abstract ideas:
“The system of claim 14, the operations further comprising: clustering client log data into candidate intents and mapping the candidate intents to the items of the matrix” (this is considered a mental process, since a person can mentally evaluate clustering client log data to candidate intents and mapping/matching these intents to items of the matrix established in claim 12, see MPEP § 2106.04(a)(2)(III)),
“creating at least one intent recommendation based on intents corresponding to the items of the training workspace” (this is considered a mental process, since a person can mentally evaluate identifying and creating a recommendation based on intents in correspondence to items from training dataset, see MPEP § 2106.04(a)(2)(III)),
If claim limitations, under their broadest reasonable interpretation, covers performance of the limitations as a mental process but for the recitation of generic compute components, then it falls within the mental process grouping of abstract ideas. Accordingly, the claim “recites” an abstract idea.
Since the claim does not recite additional elements that either integrate the judicial exception into a practical application, nor provide significantly more than the judicial exception, the claim is not patent eligible.
Claim 16:
Regarding claim 16, it is dependent upon claim 14, and thereby incorporates the limitations of, and corresponding analysis applied to claim 14. Further claim 16 recites the following abstract idea:
“The system of claim 14, wherein the comparing operation further comprises performing a check to determine is there exists an existing item with a similarity(intent_embeddingsi, item_embeddingj) that is greater than the given threshold” (this is considered a mental process, since a person can mentally evaluate using the comparing limitation from claim 12 and mentally checking if there is already an existing item with similarity that is greater than the chosen threshold, see MPEP § 2106.04(a)(2)(III)),
If claim limitations, under their broadest reasonable interpretation, covers performance of the limitations as a mental process but for the recitation of generic compute components, then it falls within the mental process grouping of abstract ideas. Accordingly, the claim “recites” an abstract idea.
Since the claim does not recite additional elements that either integrate the judicial exception into a practical application, nor provide significantly more than the judicial exception, the claim is not patent eligible.
Claim 17:
Regarding claim 17, it is dependent upon claim 14, and thereby incorporates the limitations of, and corresponding analysis applied to claim 14. Further claim 17 recites the following abstract idea:
“The system of claim 14, the operations further comprising repeating the encoding, comparing, mapping, creating the new item, and creating the matrix operations, for each workspace to update the matrix with each pair of training workspace and item” (this is considered a mental process, since a person can mentally evaluate using the encoding, comparing, mapping, creation of a new item, and creation of the matrix limitations from claim 12, to then make updates to the created matrix with each training dataset and item, see MPEP § 2106.04(a)(2)(III)),
If claim limitations, under their broadest reasonable interpretation, covers performance of the limitations as a mental process but for the recitation of generic compute components, then it falls within the mental process grouping of abstract ideas. Accordingly, the claim “recites” an abstract idea.
Since the claim does not recite additional elements that either integrate the judicial exception into a practical application, nor provide significantly more than the judicial exception, the claim is not patent eligible.
Claim 18:
Regarding claim 18, it is dependent upon claim 14, and thereby incorporates the limitations of, and corresponding analysis applied to claim 14. Further claim 18 recites the following abstract ideas:
“The system of claim 14, the operations further comprising: mapping a given client workspace into the matrix” (this is considered a mental process, since a person can mentally evaluate mapping/matching a client dataset into a matrix, see MPEP § 2106.04(a)(2)(III)),
“grouping utterances of client log data into clusters and mapping the clusters into items of the similar training workspaces using the corresponding embeddings” (this is considered a mental process, since a person can mentally evaluate grouping utterances of client log data into clusters/groupings and mapping/matching these groupings/clusters into items of the similar training datasets using the corresponding vector representations considered as embeddings, see MPEP § 2106.04(a)(2)(III)),
“recommending the first and second intents corresponding to the mapped items that exist in the similar training workspace and are absent from the client workspace” (this is considered a mental process, since a person can mentally evaluate and recommend a first and second intent in correspondence to mapped/matched items that exist from the similar training dataset and that are missing from the client dataset, see MPEP § 2106.04(a)(2)(III)),
If claim limitations, under their broadest reasonable interpretation, covers performance of the limitations as a mental process but for the recitation of generic compute components, then it falls within the mental process grouping of abstract ideas. Accordingly, the claim “recites” an abstract idea.
Further, claim 18 recites the following additional element:
“conducting a search of the training workspace that contains a similar set of items compared to the given client workspace” (In step 2A, prong 2, this is considered mere instructions to apply an exception using generic computer with the generate task-specific output operation performed by any generic computer, see MPEP § 2106.05(f)). (In step 2B, this is also considered mere instructions to apply an exception using generic computer - see MPEP § 2106.05(f)).
Since the claim does not recite additional elements that either integrate the judicial exception into a practical application, nor provide significantly more than the judicial exception, the claim is not patent eligible.
Claim 19:
Regarding claim 19, it is dependent upon claim 14, and thereby incorporates the limitations of, and corresponding analysis applied to claim 14. Further claim 19 recites the following abstract ideas:
“The system of claim 14, the operations further comprising: mapping a given client workspace into the matrix” (this is considered a mental process, since a person can mentally evaluate mapping/matching a client dataset into a matrix, see MPEP § 2106.04(a)(2)(III)),
“recommending the first and second intents corresponding to the mapped items that exist in the similar training workspace and are absent from the client workspace” (this is considered a mental process, since a person can mentally evaluate recommending first and second intents in correspondence to mapped/matched items that exist from the similar training dataset and that are missing from the client dataset, see MPEP § 2106.04(a)(2)(III)),
If claim limitations, under their broadest reasonable interpretation, covers performance of the limitations as a mental process but for the recitation of generic compute components, then it falls within the mental process grouping of abstract ideas. Accordingly, the claim “recites” an abstract idea.
Further, claim 19 recites the following additional element:
“conducting a search of the training workspace(s) that contain a similar set of items in the matrix compared to the given client workspace” (In step 2A, prong 2, this is considered mere instructions to apply an exception using generic computer with the generate task-specific output operation performed by any generic computer, see MPEP § 2106.05(f)). (In step 2B, this is also considered mere instructions to apply an exception using generic computer - see MPEP § 2106.05(f)).
Since the claim does not recite additional elements that either integrate the judicial exception into a practical application, nor provide significantly more than the judicial exception, the claim is not patent eligible.
Claim 20:
Regarding claim 20, it is dependent upon claim 14, and thereby incorporates the limitations of, and corresponding analysis applied to claim 14. Further claim 20 recites the following abstract ideas:
“The system of claim 14, the operations further comprising: grouping utterances of client log data into clusters and mapping the clusters into items of the training workspace using the corresponding embeddings” (this is considered a mental process, since a person can mentally evaluate grouping utterances of client log data into clusters/groupings and mapping/matching these groupings/clusters into items of the training datasets using the corresponding vector representations considered as embeddings, see MPEP § 2106.04(a)(2)(III)),
“recommending the first and second intents corresponding to the mapped items that exist in the training workspace” (this is considered a mental process, since a person can mentally evaluate recommending first and second intents in correspondence to mapped/matched items that exist from the training dataset, see MPEP § 2106.04(a)(2)(III)),
If claim limitations, under their broadest reasonable interpretation, covers performance of the limitations as a mental process but for the recitation of generic compute components, then it falls within the mental process grouping of abstract ideas. Accordingly, the claim “recites” an abstract idea.
Since the claim does not recite additional elements that either integrate the judicial exception into a practical application, nor provide significantly more than the judicial exception, the claim is not patent eligible.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
Claim(s) 1-4 and 10-15 are rejected under 35 U.S.C. 103 as being unpatentable over Suwandy T. et al, (US. Patent Application Publication 20210327413 A1) filed on April 4, 2020, (hereafter Suwandy), in view of Cohen W. et al, (US. Patent 8,234,285 B1) filed on July 21, 2009, (hereafter Cohen) and further in view of D'Alessandro A. et al, (US. Patent 12,260,387 B2) filed on November 2, 2022, (hereafter D'Alessandro).
Claim 1:
Regarding claim 1, Suwandy teaches “A method comprising: encoding, using at least one hardware processor, a first intent from customer provided data as an intent embedding;”
See Suwandy in paragraph [0027] where it describes “During operation, the conversational computing service may process incoming natural language user inputs that have been received via conversational entities. When a new natural language input is received, that input may be encoded as an embedding.” Here, Suwandy shows encoding a user input which can be seen as customer provided data as an embedding. Further, see Suwandy in paragraph [0028] describing “Contextual information associated with a natural language input may be taken into account by the conversational computing service in identifying an intent type and/or skill type” Here, Suwandy describes input taken in in the embodiment identifying an intent. Further see Suwandy paragraph [0100] describing, “FIG. 11 is a block diagram illustrating physical components (e.g., hardware) of a computing device 1100 with which aspects of the disclosure may be practiced. The computing device components described below may have computer executable instructions for assisting conversational entity interactions. In a basic configuration, the computing device 1100 may include at least one processing unit 1102 and a system memory 1104.”. Here, Suwandy describes a processor which is implemented in hardware, so it can be seen as a hardware processor capable of performing operations and processes of the embodiment.
Further, Suwandy teaches, “comparing, using the at least one hardware processor, the intent embedding of the first intent and an intent embedding corresponding to one or more items of a training workspace to generate a similarity score,”
See Suwandy in paragraph [0024] where it describes “The conversational computing service may maintain an embedding library that has been curated from one or more data sources. For example, the embedding library may be generated from language received from one or more general dictionaries and/or corpuses, and/or one or more domain-specific resources (e.g., subject-specific dictionaries and corpuses, technical dictionaries and corpuses, individual website domains, individual applications, website domains related to specific search criteria). The language from these sources may be processed with an encoding model to generate embeddings.” Here Suwandy establishes an embedding library which can be viewed herein to contain a training workspace. Further, see Suwandy in paragraph [0026] where it describes “Once the embedding library is generated, a developer associated with a conversational entity (e.g., bot, assistant) may add examples to the library. The examples may include identities of intent types and/or skill types corresponding to intents that a corresponding conversational entity may handle.” Here Suwandy further establishes the embedding library as a training workspace capable of generating examples which can include intents. Further, see Suwandy in paragraph [0027] where it describes “When a new natural language input is received, that input may be encoded as an embedding. In some examples, the embedding may be compressed into a binary form. The embedding may then be added to the embedding library. Once added to the embedding library, a similarity score model may be applied to the new embedding and one or more embeddings from the examples that were provided by the conversational entity developer. In some examples, the similarity score model may be a cosine model. In other examples, the similarity score model may be a Hamming model. A similarity score may be calculated between the new embedding and one or more of the example embeddings.” Here Suwandy shows generating a similarity score based on the new embedding which can be seen as the first intent to other example embeddings apart of the embedding library which can be seen as the training workspace.
Further, Suwandy teaches, “creating, using the at least one hardware processor, a matrix based on the similarity score, the created matrix including selected training workspaces”
See Suwandy in paragraph [0024] describe “As described herein, an embedding library comprises a plurality of embeddings, where words, strings, phrases, and/or sentences that have similar meaning have similar vector representations. The conversational computing service may maintain an embedding library that has been curated from one or more data sources. For example, the embedding library may be generated from language received from one or more general dictionaries and/or corpuses, and/or one or more domain-specific resources (e.g., subject-specific dictionaries and corpuses, technical dictionaries and corpuses, individual website domains, individual applications, website domains related to specific search criteria). The language from these sources may be processed with an encoding model to generate embeddings”. Here Suwandy teaches creating an embedding library with embeddings. It is well known in the art that an embedding library can be viewed as a matrix and the plurality of embeddings can be seen as training workspaces. Further, Suwandy in paragraph [0027] describes “When a new natural language input is received, that input may be encoded as an embedding. In some examples, the embedding may be compressed into a binary form. The embedding may then be added to the embedding library. Once added to the embedding library, a similarity score model may be applied to the new embedding and one or more embeddings from the examples that were provided by the conversational entity developer. In some examples, the similarity score model may be a cosine model. In other examples, the similarity score model may be a Hamming model. A similarity score may be calculated between the new embedding and one or more of the example embeddings.” Here Suwandy teaches the embedding library which can be seen as a created matrix being based on a similarity score between a new embedding and one or more previous embeddings which can be seen as training workspaces.
Further, Suwandy teaches, “training, using the at least one hardware processor, at least a first machine learning model using one or more of the selected training workspaces of the created matrix;”
See Suwandy in paragraph [0056] where it describes “FIG. 3A illustrates a simplified graph 300A of sentence embeddings in an embedding library and the training of a language model via inclusion of two new examples to the embedding library.” Here, Suwandy teaches training a language model, using two examples of the embedding library, as mentioned before the example embeddings of the embedding library can be viewed as training workspaces and using two example embeddings from the embedding library can be seen as selecting one or more training workspaces. As understood in the art an embedding library can be seen as a created matrix as well.
However, Suwandy did not explicitly teach “mapping, using the at least one hardware processor, the first intent to a similar item of the one or more items and incrementing a corresponding count of the similar item by one in response to the similarity score being greater than a given threshold; training, using the at least one hardware processor, at least a second machine learning model using the created matrix; and facilitating, using the at least one hardware processor, deployment of the at least second machine learning model for performing inferencing.”
In the same field of art, Cohen teaches, “mapping, using the at least one hardware processor, the first intent to a similar item of the one or more items and incrementing a corresponding count of the similar item by one in response to the similarity score being greater than a given threshold;”
See Cohen in column 1 lines 29-37 describes “A common step in integrating heterogeneous datasets is determining a mapping between objects from one dataset and objects from another dataset. This step is often referred to as record linkage, matching, and/or de-duping. One useful matching strategy is to use a threshold similarity function that generates a similarity score from the feature values and identifies objects as identical if the similarity score exceeds a threshold value.” Here Cohen teaches mapping one object to other objects of another dataset if the similarity score exceeds a threshold. In an analogous system we can see the mapping of one object to other objects as mapping an intent and a similar item of one or more items. Further, see Cohen in column 12 lines 25-30 describe “For each feature value and context similarity value pair in the context similarity list generated by the process 400, the process 600 increments counter a by CX, increments counter b by CX2, and increments counter c by 1 (604). For example, the contextual similarity engine 112 can increment the counters for each (f, CX) pair.” Here Cohen then teaches incrementing a count based on the mapping corresponding to the similarity score which was earlier established to be determined by exceeding or being greater than a threshold.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the base reference of Suwandy with the teachings of Cohen by using Suwandy’s teachings of converting an intent into and embedding, comparing the embedding, creating a matrix based on a similarity score, and training a model, and incorporate with Cohen’s teachings of mapping items to a similar and incrementing a corresponding count of the similar.
One of ordinary skill in the art would be motivated to do so because by integrating Cohen’s frameworks into the methods of Suwandy, which are both in relation to context analysis and similarity measurement, one of ordinary skill in the art would bring “selecting, by a data processing apparatus, object representations from a dataset storing a plurality of object representations, each object representation being an association of an object identifier that identifies an object instance in the dataset and corresponds to an object, a context value that identifies a context of the object, and a set of feature values that identify features of the object, and wherein each object identifier is unique in the dataset” (Cohen, column 2 lines 5-13).
However, Suwandy in view of Cohen did not explicitly teach “training, using the at least one hardware processor, at least a second machine learning model using the created matrix; facilitating, using the at least one hardware processor, deployment of the at least second machine learning model for performing inferencing.”
Further, D’Alessandro in the same field of art teaches, “training, using the at least one hardware processor, at least a second machine learning model using the created matrix”
See D’Alessandro in Column 5 lines 6-8 where it describes “In some examples, the system includes a machine learning (ML) engine with one or more ML models, which the system may train using training data.” Here, D’Alessandro establishes a ML engine comprising one or more ML models which can be trained using training data which as mentioned before can be seen as training workspaces, the analogous system already teaches a created matrix using training data. Further see D’Alessandro in Column 5 lines 44-47 where it describes “In some examples, the system uses the first trained ML engine to identify the intent for the transaction and uses a second trained ML engine to generate the recommended transaction.” Here D’Alessandro establishes a second trained ML engine and in the analogous system because a ML engine here is trained using training data and it is taught to already create a matrix comprising of training workspaces, we can view this as training a second ML model using a created matrix.
Further, D’Alessandro teaches, “facilitating, using the at least one hardware processor, deployment of the at least second machine learning model for performing inferencing.”
See D’Alessandro in Column 5 lines 44-47 where it describes “In some examples, the system uses the first trained ML engine to identify the intent for the transaction and uses a second trained ML engine to generate the recommended transaction.” Here, D’Alessandro shows using the second machine learning (ML) engine which is using at least another ML model aside from the first ML engine can be used to generate a recommendation which can be seen as inferencing.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the base reference of Suwandy with the teachings of Cohen and D’Alessandro by using Suwandy’s teachings of converting an intent into and embedding, comparing the embedding and training a model, and incorporate with Cohen and D’Alessandro’s teachings of mapping items to a similar and incrementing a corresponding count of the similar, creating a matrix based on a similarity score, and using a second trained model to perform inferencing.
One of ordinary skill in the art would be motivated to do so because by integrating D’Alessandro’s frameworks into the methods of Suwandy and Cohen, one with ordinary skill in the art would “provide customized recommendations (e.g., recommended transactions) that are customized and/or tailored specifically to users based on their histories, their account information, intents determined behind their transaction(s), or combinations thereof. This improves over systems that are unable to provide recommendations, or provide standardized recommendations without such customization.” (D’Alessandro, column 6 lines 35-41).
Claim 2:
Regarding claim 2, Suwandy in view of Cohen and D’Alessandro teaches the limitations in claim 1.
Further, D’Alessandro teaches “method of claim 1, further comprising performing inferencing using the deployed at least second trained machine learning model.”
See D’Alessandro in Column 5 lines 46-47 where it describes “In some examples, the system uses the first trained ML engine to identify the intent for the transaction and uses a second trained ML engine to generate the recommended transaction.” Here, D’Alessandro shows using the second machine learning (ML) engine, which is using at least another ML model aside from the first ML engine, to generate a recommendation which can be seen as performing the inferencing.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the base reference of Suwandy with the teachings of Cohen and D’Alessandro by using Suwandy’s teachings of converting an intent into and embedding, comparing the embedding and training a model, and incorporate with Cohen and D’Alessandro’s teachings of mapping items to a similar and incrementing a corresponding count of the similar, and using a second trained model to perform inferencing.
One of ordinary skill in the art would be motivated to do so because by integrating D’Alessandro’s frameworks into the methods of Suwandy and Cohen, one with ordinary skill in the art would “provide customized recommendations (e.g., recommended transactions) that are customized and/or tailored specifically to users based on their histories, their account information, intents determined behind their transaction(s), or combinations thereof. This improves over systems that are unable to provide recommendations, or provide standardized recommendations without such customization.” (D’Alessandro, column 6 lines 35-41).
Claim 3:
Regarding claim 3, Suwandy in view of Cohen and D’Alessandro teaches the limitations in claim 1.
Further, Suwandy teaches “method of claim 1, further comprising repeating the comparing operation and creating a new item corresponding to a second intent in response to the similarity score being less than the given threshold”
See Suwandy in paragraph [0027] where it describes “A similarity score may be calculated between the new embedding and one or more of the example embeddings. A K Nearest Neighbors (KNN) model may then be utilized to identify one or more intent types and/or skill types that correspond to the new embedding. In some examples, a nearest neighbor score value may be calculated for each of the one or more intent types and/or skill types based on application of the KNN model.” Here, Suwandy teaches repeating a comparing operation to attempt to identify one or more intents. Further, see Suwandy in paragraph [0033] where it describes “In some examples, if there is no intent type and/or skill type for which a similarity score exceeds a threshold value, a determination may be made that there is an unknown intent.” Here, Suwandy teaches a new item being created in response to a similarity score not exceeding a threshold making it be less than a threshold, this new item being an unknown intent which is also the second intent.
Further, Cohen teaches “and setting a count corresponding to the new item to one.”
See Cohen in column 12 lines 25-30 where it describes “For each feature value and context similarity value pair in the context similarity list generated by the process 400, the process 600 increments counter a by CX, increments counter b by CX2, and increments counter c by 1 (604). For example, the contextual similarity engine 112 can increment the counters for each (f, CX) pair. Here, Cohen teaches incrementing a count of a corresponding item to one. In an analogous system one could perform this operation in combination with repeating the comparing operation.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the base reference of Suwandy with the teachings of Cohen by using Suwandy’s teachings of converting an intent into and embedding, comparing the embedding, creating a matrix based on a similarity score, and training a model, and incorporate with Cohen’s teachings of mapping items to a similar and incrementing a corresponding count of the similar.
One of ordinary skill in the art would be motivated to do so because by integrating Cohen’s frameworks into the methods of Suwandy, which are both in relation to context analysis and similarity measurement, one of ordinary skill in the art would bring “selecting, by a data processing apparatus, object representations from a dataset storing a plurality of object representations, each object representation being an association of an object identifier that identifies an object instance in the dataset and corresponds to an object, a context value that identifies a context of the object, and a set of feature values that identify features of the object, and wherein each object identifier is unique in the dataset” (Cohen, column 2 lines 5-13).
Claim 4:
Regarding claim 4, Suwandy in view of Cohen and D’Alessandro teaches the limitations in claim 3.
Further, Suwandy teaches “method of claim 3, further comprising: clustering client log data into candidate intents and mapping the candidate intents to the items of the matrix”
See Suwandy paragraph [0026] describing, “Once the embedding library is generated, a developer associated with a conversational entity (e.g., bot, assistant) may add examples to the library. The examples may include identities of intent types and/or skill types corresponding to intents that a corresponding conversational entity may handle. As described herein, an intent comprises a classification of one or more inputs that may be received by a conversational bot, such as into a type of skill or action that may be performed by the conversational bot in response to the input.” Here, Suwandy establishes that the example embeddings can include intents. Further, see Suwandy paragraph [0027] describing, “During operation, the conversational computing service may process incoming natural language user inputs that have been received via conversational entities. When a new natural language input is received, that input may be encoded as an embedding. In some examples, the embedding may be compressed into a binary form. The embedding may then be added to the embedding library. Once added to the embedding library, a similarity score model may be applied to the new embedding and one or more embeddings from the examples that were provided by the conversational entity developer. In some examples, the similarity score model may be a cosine model. In other examples, the similarity score model may be a Hamming model. A similarity score may be calculated between the new embedding and one or more of the example embeddings. A K Nearest Neighbors (KNN) model may then be utilized to identify one or more intent types and/or skill types that correspond to the new embedding.” Here, Suwandy teaches grouping a new embedding which can be seen as an intent from a user input which can be considered client log data into candidate intents by identifying one or more intent types that correspond to the new embedding. The identification of these intents can be seen as mapping the intents to items of the embedding library, which has been established to be considered a matrix, because these intent types come from example embeddings which are items of this matrix.
Further, Suwandy teaches “creating at least one intent recommendation based on intents corresponding to the items of the training workspace”
See Suwandy paragraph [0027] describing, “In some examples, a nearest neighbor score value may be calculated for each of the one or more intent types and/or skill types based on application of the KNN model. In some examples, if a nearest neighbor score value for an intent type and/or skill type exceeds a threshold value, a response or action corresponding to that intent and/or skill type may be performed by the conversational entity.” Here, Suwandy teaches performing an action using a conversational entity based on an intent being identified. Further See Suwandy paragraph [0027] describing, “In some examples, a nearest neighbor score value may be calculated for each of the one or more intent types and/or skill types based on application of the KNN model. In some examples, if a nearest neighbor score value for an intent type and/or skill type exceeds a threshold value, a response or action corresponding to that intent and/or skill type may be performed by the conversational entity.” Here, Suwandy teaches performing an action or response using a conversational entity based on an intent being identified. Which can be seen as creating an intent recommendation.
Claim 10:
Regarding claim 10, Suwandy in view of Cohen and D’Alessandro teaches the limitations in claim 2.
Further, Suwandy teaches “method of claim 2, wherein the inferencing is performed to generate a recommendation for a conversational artificial intelligence task”
See Suwandy in paragraph [0028] where it describes “Contextual information associated with a natural language input may be taken into account by the conversational computing service in identifying an intent type and/or skill type.” Here, Suwandy describes taking in an input to identify an intent using a conversational computing service which can be seen as generating a recommendation for a conversational task. Further, see Suwandy in paragraph [0021] where it describes “Examples of the disclosure provide systems, methods, and devices for training language models that may be utilized in enabling conversational computing communications. A conversational entity, such as a conversational bot and/or a conversational assistant, may be published by, and/or associated with, a particular group, institution, or person.” Here, Suwandy shows the method being used to perform conversational tasks which because they are performed by bots and/or assistants they are conversational artificial intelligence tasks being perfromed.
Claim 11:
Regarding claim 11, Suwandy teaches “A computer program product, comprising: one or more tangible computer-readable storage media and program instructions stored on at least one of the one or more tangible computer-readable storage media, the program instructions executable by a processor, the program instructions comprising: encoding a first intent from customer provided data as an intent embedding;”
See Suwandy in paragraph [0027] where it describes “During operation, the conversational computing service may process incoming natural language user inputs that have been received via conversational entities. When a new natural language input is received, that input may be encoded as an embedding.” Here, Suwandy shows encoding a user input which can be seen as customer provided data as an embedding. Further, see Suwandy in paragraph [0028] describing “Contextual information associated with a natural language input may be taken into account by the conversational computing service in identifying an intent type and/or skill type” Here, Suwandy describes input taken in in the embodiment identifying an intent. Further see Suwandy paragraph [0100] describing, “FIG. 11 is a block diagram illustrating physical components (e.g., hardware) of a computing device 1100 with which aspects of the disclosure may be practiced. The computing device components described below may have computer executable instructions for assisting conversational entity interactions. In a basic configuration, the computing device 1100 may include at least one processing unit 1102 and a system memory 1104.”. Here, Suwandy describes a processor which is implemented in hardware, so it can be seen as a hardware processor capable of performing operations and processes of the embodiment. Further, see Suwandy in paragraph [0104] describing, “The term computer readable media as used herein may include computer storage media. Computer storage media may include volatile and nonvolatile, removable and non-removable media implemented in any method or technology for storage of information, such as computer readable instructions, data structures, or program modules. The system memory 1104, the removable storage device 1109, and the non-removable storage device 1110 are all computer storage media examples (e.g., memory storage). Computer storage media may include RAM, ROM, electrically erasable read-only memory (EEPROM), flash memory or other memory technology, CD-ROM, digital versatile disks (DVD) or other optical storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other article of manufacture which can be used to store information and which can be accessed by the computing device 1100.” This shows the mentioned computing device 1100 using a tangible computer-readable storage media to further do operations described.
Further, Suwandy teaches, “comparing the intent embedding of the first intent and an intent embedding corresponding to one or more items of a training workspace to generate a similarity score,”
See Suwandy in paragraph [0024] where it describes “The conversational computing service may maintain an embedding library that has been curated from one or more data sources. For example, the embedding library may be generated from language received from one or more general dictionaries and/or corpuses, and/or one or more domain-specific resources (e.g., subject-specific dictionaries and corpuses, technical dictionaries and corpuses, individual website domains, individual applications, website domains related to specific search criteria). The language from these sources may be processed with an encoding model to generate embeddings.” Here Suwandy establishes an embedding library which can be viewed herein to contain a training workspace. Further, see Suwandy in paragraph [0026] where it describes “Once the embedding library is generated, a developer associated with a conversational entity (e.g., bot, assistant) may add examples to the library. The examples may include identities of intent types and/or skill types corresponding to intents that a corresponding conversational entity may handle.” Here Suwandy further establishes the embedding library as a training workspace capable of generating examples which can include intents. Further, see Suwandy in paragraph [0027] where it describes “When a new natural language input is received, that input may be encoded as an embedding. In some examples, the embedding may be compressed into a binary form. The embedding may then be added to the embedding library. Once added to the embedding library, a similarity score model may be applied to the new embedding and one or more embeddings from the examples that were provided by the conversational entity developer. In some examples, the similarity score model may be a cosine model. In other examples, the similarity score model may be a Hamming model. A similarity score may be calculated between the new embedding and one or more of the example embeddings.” Here Suwandy shows generating a similarity score based on the new embedding which can be seen as the first intent to other example embeddings apart of the embedding library which can be seen as the training workspace.
Further, Suwandy teaches, “creating a matrix based on the similarity score, the created matrix including selected training workspaces”
See Suwandy in paragraph [0024] describe “As described herein, an embedding library comprises a plurality of embeddings, where words, strings, phrases, and/or sentences that have similar meaning have similar vector representations. The conversational computing service may maintain an embedding library that has been curated from one or more data sources. For example, the embedding library may be generated from language received from one or more general dictionaries and/or corpuses, and/or one or more domain-specific resources (e.g., subject-specific dictionaries and corpuses, technical dictionaries and corpuses, individual website domains, individual applications, website domains related to specific search criteria). The language from these sources may be processed with an encoding model to generate embeddings”. Here Suwandy teaches creating an embedding library with embeddings. It is well known in the art that an embedding library can be viewed as a matrix and the plurality of embeddings can be seen as training workspaces. Further, Suwandy in paragraph [0027] describes “When a new natural language input is received, that input may be encoded as an embedding. In some examples, the embedding may be compressed into a binary form. The embedding may then be added to the embedding library. Once added to the embedding library, a similarity score model may be applied to the new embedding and one or more embeddings from the examples that were provided by the conversational entity developer. In some examples, the similarity score model may be a cosine model. In other examples, the similarity score model may be a Hamming model. A similarity score may be calculated between the new embedding and one or more of the example embeddings.” Here Suwandy teaches the embedding library which can be seen as a created matrix being based on a similarity score between a new embedding and one or more previous embeddings which can be seen as training workspaces.
Further, Suwandy teaches, “training at least a first machine learning model using one or more of the selected training workspaces of the created matrix;
See Suwandy in paragraph [0056] where it describes “FIG. 3A illustrates a simplified graph 300A of sentence embeddings in an embedding library and the training of a language model via inclusion of two new examples to the embedding library.” Here, Suwandy teaches training a language model, using two examples of the embedding library, as mentioned before the example embeddings of the embedding library can be viewed as training workspaces and using two example embeddings from the embedding library can be seen as selecting one or more training workspaces. As understood in the art an embedding library can be seen as a created matrix as well.
However, Suwandy did not explicitly teach “mapping, the first intent to a similar item of the one or more items and incrementing a corresponding count of the similar item by one in response to the similarity score being greater than a given threshold; training, using the at least one hardware processor, at least a second machine learning model using the created matrix; and facilitating deployment of the at least second machine learning model for performing inferencing.”
In the same field of art, Cohen teaches, “mapping the first intent to a similar item of the one or more items and incrementing a corresponding count of the similar item by one in response to the similarity score being greater than a given threshold;”
See Cohen in column 1 lines 29-37 describes “A common step in integrating heterogeneous datasets is determining a mapping between objects from one dataset and objects from another dataset. This step is often referred to as record linkage, matching, and/or de-duping. One useful matching strategy is to use a threshold similarity function that generates a similarity score from the feature values and identifies objects as identical if the similarity score exceeds a threshold value.” Here Cohen teaches mapping one object to other objects of another dataset if the similarity score exceeds a threshold. In an analogous system we can see the mapping of one object to other objects as mapping an intent and a similar item of one or more items. Further, see Cohen in column 12 lines 25-30 describe “For each feature value and context similarity value pair in the context similarity list generated by the process 400, the process 600 increments counter a by CX, increments counter b by CX2, and increments counter c by 1 (604). For example, the contextual similarity engine 112 can increment the counters for each (f, CX) pair.” Here Cohen then teaches incrementing a count based on the mapping corresponding to the similarity score which was earlier established to be determined by exceeding or being greater than a threshold.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the base reference of Suwandy with the teachings of Cohen by using Suwandy’s teachings of converting an intent into and embedding, comparing the embedding, creating a matrix based on a similarity score, and training a model, and incorporate with Cohen’s teachings of mapping items to a similar and incrementing a corresponding count of the similar.
One of ordinary skill in the art would be motivated to do so because by integrating Cohen’s frameworks into the methods of Suwandy, which are both in relation to context analysis and similarity measurement, one of ordinary skill in the art would bring “selecting, by a data processing apparatus, object representations from a dataset storing a plurality of object representations, each object representation being an association of an object identifier that identifies an object instance in the dataset and corresponds to an object, a context value that identifies a context of the object, and a set of feature values that identify features of the object, and wherein each object identifier is unique in the dataset” (Cohen, column 2 lines 5-13).
However, Suwandy in view of Cohen did not explicitly teach “training, using the at least one hardware processor, at least a second machine learning model using the created matrix; facilitating deployment of the at least second machine learning model for performing inferencing.”
Further, D’Alessandro in the same field of art teaches, “training, using the at least one hardware processor, at least a second machine learning model using the created matrix”
See D’Alessandro in Column 5 lines 6-8 where it describes “In some examples, the system includes a machine learning (ML) engine with one or more ML models, which the system may train using training data.” Here, D’Alessandro establishes a ML engine comprising one or more ML models which can be trained using training data which as mentioned before can be seen as training workspaces, the analogous system already teaches a created matrix using training data. Further see D’Alessandro in Column 5 lines 44-47 where it describes “In some examples, the system uses the first trained ML engine to identify the intent for the transaction and uses a second trained ML engine to generate the recommended transaction.” Here D’Alessandro establishes a second trained ML engine and in the analogous system because a ML engine here is trained using training data and it is taught to already create a matrix comprising of training workspaces, we can view this as training a second ML model using a created matrix.
Further, D’Alessandro teaches, “facilitating deployment of the at least second machine learning model for performing inferencing.”
See D’Alessandro in Column 5 lines 44-47 where it describes “In some examples, the system uses the first trained ML engine to identify the intent for the transaction and uses a second trained ML engine to generate the recommended transaction.” Here, D’Alessandro shows using the second machine learning (ML) engine which is using at least another ML model aside from the first ML engine can be used to generate a recommendation which can be seen as inferencing.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the base reference of Suwandy with the teachings of Cohen and D’Alessandro by using Suwandy’s teachings of converting an intent into and embedding, comparing the embedding and training a model, and incorporate with Cohen and D’Alessandro’s teachings of mapping items to a similar and incrementing a corresponding count of the similar, creating a matrix based on a similarity score, and using a second trained model to perform inferencing.
One of ordinary skill in the art would be motivated to do so because by integrating D’Alessandro’s frameworks into the methods of Suwandy and Cohen, one with ordinary skill in the art would “provide customized recommendations (e.g., recommended transactions) that are customized and/or tailored specifically to users based on their histories, their account information, intents determined behind their transaction(s), or combinations thereof. This improves over systems that are unable to provide recommendations, or provide standardized recommendations without such customization.” (D’Alessandro, column 6 lines 35-41).
Claim 12:
Regarding claim 12, Suwandy teaches “A system comprising: a memory; and at least one processor, coupled to said memory, and operative to perform operations comprising: encoding a first intent from customer provided data as an intent embedding;”
See Suwandy in paragraph [0027] where it describes “During operation, the conversational computing service may process incoming natural language user inputs that have been received via conversational entities. When a new natural language input is received, that input may be encoded as an embedding.” Here, Suwandy shows encoding a user input which can be seen as customer provided data as an embedding. Further, see Suwandy in paragraph [0028] describing “Contextual information associated with a natural language input may be taken into account by the conversational computing service in identifying an intent type and/or skill type” Here, Suwandy describes input taken in in the embodiment identifying an intent. Further see Suwandy paragraph [0100] describing, “FIG. 11 is a block diagram illustrating physical components (e.g., hardware) of a computing device 1100 with which aspects of the disclosure may be practiced. The computing device components described below may have computer executable instructions for assisting conversational entity interactions. In a basic configuration, the computing device 1100 may include at least one processing unit 1102 and a system memory 1104.”. Here, Suwandy describes a processor which is implemented in hardware, so it can be seen as a hardware processor with a memory coupled to it capable of performing operations and processes of the embodiment.
Further, Suwandy teaches, “comparing the intent embedding of the first intent and an intent embedding corresponding to one or more items of a training workspace to generate a similarity score,”
See Suwandy in paragraph [0024] where it describes “The conversational computing service may maintain an embedding library that has been curated from one or more data sources. For example, the embedding library may be generated from language received from one or more general dictionaries and/or corpuses, and/or one or more domain-specific resources (e.g., subject-specific dictionaries and corpuses, technical dictionaries and corpuses, individual website domains, individual applications, website domains related to specific search criteria). The language from these sources may be processed with an encoding model to generate embeddings.” Here Suwandy establishes an embedding library which can be viewed herein to contain a training workspace. Further, see Suwandy in paragraph [0026] where it describes “Once the embedding library is generated, a developer associated with a conversational entity (e.g., bot, assistant) may add examples to the library. The examples may include identities of intent types and/or skill types corresponding to intents that a corresponding conversational entity may handle.” Here Suwandy further establishes the embedding library as a training workspace capable of generating examples which can include intents. Further, see Suwandy in paragraph [0027] where it describes “When a new natural language input is received, that input may be encoded as an embedding. In some examples, the embedding may be compressed into a binary form. The embedding may then be added to the embedding library. Once added to the embedding library, a similarity score model may be applied to the new embedding and one or more embeddings from the examples that were provided by the conversational entity developer. In some examples, the similarity score model may be a cosine model. In other examples, the similarity score model may be a Hamming model. A similarity score may be calculated between the new embedding and one or more of the example embeddings.” Here Suwandy shows generating a similarity score based on the new embedding which can be seen as the first intent to other example embeddings apart of the embedding library which can be seen as the training workspace.
Further, Suwandy teaches, “creating a matrix based on the similarity score, the created matrix including selected training workspaces”
See Suwandy in paragraph [0024] describe “As described herein, an embedding library comprises a plurality of embeddings, where words, strings, phrases, and/or sentences that have similar meaning have similar vector representations. The conversational computing service may maintain an embedding library that has been curated from one or more data sources. For example, the embedding library may be generated from language received from one or more general dictionaries and/or corpuses, and/or one or more domain-specific resources (e.g., subject-specific dictionaries and corpuses, technical dictionaries and corpuses, individual website domains, individual applications, website domains related to specific search criteria). The language from these sources may be processed with an encoding model to generate embeddings”. Here Suwandy teaches creating an embedding library with embeddings. It is well known in the art that an embedding library can be viewed as a matrix and the plurality of embeddings can be seen as training workspaces. Further, Suwandy in paragraph [0027] describes “When a new natural language input is received, that input may be encoded as an embedding. In some examples, the embedding may be compressed into a binary form. The embedding may then be added to the embedding library. Once added to the embedding library, a similarity score model may be applied to the new embedding and one or more embeddings from the examples that were provided by the conversational entity developer. In some examples, the similarity score model may be a cosine model. In other examples, the similarity score model may be a Hamming model. A similarity score may be calculated between the new embedding and one or more of the example embeddings.” Here Suwandy teaches the embedding library which can be seen as a created matrix being based on a similarity score between a new embedding and one or more previous embeddings which can be seen as training workspaces.
Further, Suwandy teaches, “training at least a first machine learning model using one or more of the selected training workspaces of the created matrix;
See Suwandy in paragraph [0056] where it describes “FIG. 3A illustrates a simplified graph 300A of sentence embeddings in an embedding library and the training of a language model via inclusion of two new examples to the embedding library.” Here, Suwandy teaches training a language model, using two examples of the embedding library, as mentioned before the example embeddings of the embedding library can be viewed as training workspaces and using two example embeddings from the embedding library can be seen as selecting one or more training workspaces. As understood in the art an embedding library can be seen as a created matrix as well.
However, Suwandy did not explicitly teach “mapping, the first intent to a similar item of the one or more items and incrementing a corresponding count of the similar item by one in response to the similarity score being greater than a given threshold; training at least a second machine learning model using the created matrix; and facilitating deployment of the at least second machine learning model for performing inferencing.”
In the same field of art, Cohen teaches, “mapping the first intent to a similar item of the one or more items and incrementing a corresponding count of the similar item by one in response to the similarity score being greater than a given threshold;”
See Cohen in column 1 lines 29-37 describes “A common step in integrating heterogeneous datasets is determining a mapping between objects from one dataset and objects from another dataset. This step is often referred to as record linkage, matching, and/or de-duping. One useful matching strategy is to use a threshold similarity function that generates a similarity score from the feature values and identifies objects as identical if the similarity score exceeds a threshold value.” Here Cohen teaches mapping one object to other objects of another dataset if the similarity score exceeds a threshold. In an analogous system we can see the mapping of one object to other objects as mapping an intent and a similar item of one or more items. Further, see Cohen in column 12 lines 25-30 describe “For each feature value and context similarity value pair in the context similarity list generated by the process 400, the process 600 increments counter a by CX, increments counter b by CX2, and increments counter c by 1 (604). For example, the contextual similarity engine 112 can increment the counters for each (f, CX) pair.” Here Cohen then teaches incrementing a count based on the mapping corresponding to the similarity score which was earlier established to be determined by exceeding or being greater than a threshold.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the base reference of Suwandy with the teachings of Cohen by using Suwandy’s teachings of converting an intent into and embedding, comparing the embedding, creating a matrix based on a similarity score, and training a model, and incorporate with Cohen’s teachings of mapping items to a similar and incrementing a corresponding count of the similar.
One of ordinary skill in the art would be motivated to do so because by integrating Cohen’s frameworks into the methods of Suwandy, which are both in relation to context analysis and similarity measurement, one of ordinary skill in the art would bring “selecting, by a data processing apparatus, object representations from a dataset storing a plurality of object representations, each object representation being an association of an object identifier that identifies an object instance in the dataset and corresponds to an object, a context value that identifies a context of the object, and a set of feature values that identify features of the object, and wherein each object identifier is unique in the dataset” (Cohen, column 2 lines 5-13).
However, Suwandy in view of Cohen did not explicitly teach “training at least a second machine learning model using the created matrix; facilitating deployment of the at least second machine learning model for performing inferencing.”
Further, D’Alessandro in the same field of art teaches, “training at least a second machine learning model using the created matrix”
See D’Alessandro in Column 5 lines 6-8 where it describes “In some examples, the system includes a machine learning (ML) engine with one or more ML models, which the system may train using training data.” Here, D’Alessandro establishes a ML engine comprising one or more ML models which can be trained using training data which as mentioned before can be seen as training workspaces, the analogous system already teaches a created matrix using training data. Further see D’Alessandro in Column 5 lines 44-47 where it describes “In some examples, the system uses the first trained ML engine to identify the intent for the transaction and uses a second trained ML engine to generate the recommended transaction.” Here D’Alessandro establishes a second trained ML engine and in the analogous system because a ML engine here is trained using training data and it is taught to already create a matrix comprising of training workspaces, we can view this as training a second ML model using a created matrix.
Further, D’Alessandro teaches, “facilitating deployment of the at least second machine learning model for performing inferencing.”
See D’Alessandro in Column 5 lines 44-47 where it describes “In some examples, the system uses the first trained ML engine to identify the intent for the transaction and uses a second trained ML engine to generate the recommended transaction.” Here, D’Alessandro shows using the second machine learning (ML) engine which is using at least another ML model aside from the first ML engine can be used to generate a recommendation which can be seen as inferencing.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the base reference of Suwandy with the teachings of Cohen and D’Alessandro by using Suwandy’s teachings of converting an intent into and embedding, comparing the embedding and training a model, and incorporate with Cohen and D’Alessandro’s teachings of mapping items to a similar and incrementing a corresponding count of the similar, creating a matrix based on a similarity score, and using a second trained model to perform inferencing.
One of ordinary skill in the art would be motivated to do so because by integrating D’Alessandro’s frameworks into the methods of Suwandy and Cohen, one with ordinary skill in the art would “provide customized recommendations (e.g., recommended transactions) that are customized and/or tailored specifically to users based on their histories, their account information, intents determined behind their transaction(s), or combinations thereof. This improves over systems that are unable to provide recommendations, or provide standardized recommendations without such customization.” (D’Alessandro, column 6 lines 35-41).
Claim 13:
Regarding claim 13, Suwandy in view of D’Alessandro and Cohen teaches the limitations in claim 12.
Further, D’Alessandro teaches “The system of claim 12, the operations further comprising performing inferencing using the deployed at least second trained machine learning model.”
See D’Alessandro in Column 5 lines 46-47 where it describes “In some examples, the system uses the first trained ML engine to identify the intent for the transaction and uses a second trained ML engine to generate the recommended transaction.” Here, D’Alessandro shows using the second machine learning (ML) engine, which is using at least another ML model aside from the first ML engine, to generate a recommendation which can be seen as performing the inferencing.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the base reference of Suwandy with the teachings of Cohen and D’Alessandro by using Suwandy’s teachings of converting an intent into and embedding, comparing the embedding and training a model, and incorporate with Cohen and D’Alessandro’s teachings of mapping items to a similar and incrementing a corresponding count of the similar, creating a matrix based on a similarity score, and using a second trained model to perform inferencing.
One of ordinary skill in the art would be motivated to do so because by integrating D’Alessandro’s frameworks into the methods of Suwandy and Cohen, one with ordinary skill in the art would “provide customized recommendations (e.g., recommended transactions) that are customized and/or tailored specifically to users based on their histories, their account information, intents determined behind their transaction(s), or combinations thereof. This improves over systems that are unable to provide recommendations, or provide standardized recommendations without such customization.” (D’Alessandro, column 6 lines 35-41).
Claim 14:
Regarding claim 14, Suwandy in view of Cohen and D’Alessandro teaches the limitations in claim 12.
Further, Suwandy teaches “The system of claim 12, the operations further comprising repeating the comparing operation and creating a new item corresponding to a second intent in response to the similarity score being less than the given threshold”
See Suwandy in paragraph [0027] where it describes “A similarity score may be calculated between the new embedding and one or more of the example embeddings. A K Nearest Neighbors (KNN) model may then be utilized to identify one or more intent types and/or skill types that correspond to the new embedding. In some examples, a nearest neighbor score value may be calculated for each of the one or more intent types and/or skill types based on application of the KNN model.” Here, Suwandy teaches repeating a comparing operation to attempt to identify one or more intents. Further, see Suwandy in paragraph [0033] where it describes “In some examples, if there is no intent type and/or skill type for which a similarity score exceeds a threshold value, a determination may be made that there is an unknown intent.” Here, Suwandy teaches a new item being created in response to a similarity score not exceeding a threshold making it be less than a threshold, this new item being an unknown intent which is also the second intent.
Further, Cohen teaches “and setting a count corresponding to the new item to one.”
See Cohen in column 12 lines 25-30 where it describes “For each feature value and context similarity value pair in the context similarity list generated by the process 400, the process 600 increments counter a by CX, increments counter b by CX2 and increments counter c by 1 (604). For example, the contextual similarity engine 112 can increment the counters for each (f, CX) pair. Here, Cohen teaches incrementing a count of a corresponding item to one. In an analogous system one could perform this operation in combination with repeating the comparing operation.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the base reference of Suwandy with the teachings of Cohen by using Suwandy’s teachings of converting an intent into and embedding, comparing the embedding, creating a matrix based on a similarity score, and training a model, and incorporate with Cohen’s teachings of mapping items to a similar and incrementing a corresponding count of the similar.
One of ordinary skill in the art would be motivated to do so because by integrating Cohen’s frameworks into the methods of Suwandy, which are both in relation to context analysis and similarity measurement, one of ordinary skill in the art would bring “selecting, by a data processing apparatus, object representations from a dataset storing a plurality of object representations, each object representation being an association of an object identifier that identifies an object instance in the dataset and corresponds to an object, a context value that identifies a context of the object, and a set of feature values that identify features of the object, and wherein each object identifier is unique in the dataset” (Cohen, column 2 lines 5-13).
Claim 15:
Regarding claim 15, Suwandy in view of Cohen and D’Alessandro teaches the limitations in claim 14.
Further, Suwandy teaches “The system of claim 14, the operations further comprising: clustering client log data into candidate intents and mapping the candidate intents to the items of the matrix”
See Suwandy paragraph [0026] describing, “Once the embedding library is generated, a developer associated with a conversational entity (e.g., bot, assistant) may add examples to the library. The examples may include identities of intent types and/or skill types corresponding to intents that a corresponding conversational entity may handle. As described herein, an intent comprises a classification of one or more inputs that may be received by a conversational bot, such as into a type of skill or action that may be performed by the conversational bot in response to the input.” Here, Suwandy establishes that the example embeddings can include intents. Further, see Suwandy paragraph [0027] describing, “During operation, the conversational computing service may process incoming natural language user inputs that have been received via conversational entities. When a new natural language input is received, that input may be encoded as an embedding. In some examples, the embedding may be compressed into a binary form. The embedding may then be added to the embedding library. Once added to the embedding library, a similarity score model may be applied to the new embedding and one or more embeddings from the examples that were provided by the conversational entity developer. In some examples, the similarity score model may be a cosine model. In other examples, the similarity score model may be a Hamming model. A similarity score may be calculated between the new embedding and one or more of the example embeddings. A K Nearest Neighbors (KNN) model may then be utilized to identify one or more intent types and/or skill types that correspond to the new embedding.” Here, Suwandy teaches grouping a new embedding which can be seen as an intent from a user input which can be considered client log data into candidate intents by identifying one or more intent types that correspond to the new embedding. The identification of these intents can be seen as mapping the intents to items of the embedding library, which has been established to be considered a matrix, because these intent types come from example embeddings which are items of this matrix.
Further, Suwandy teaches “creating at least one intent recommendation based on intents corresponding to the items of the training workspace”
See Suwandy paragraph [0027] describing, “In some examples, a nearest neighbor score value may be calculated for each of the one or more intent types and/or skill types based on application of the KNN model. In some examples, if a nearest neighbor score value for an intent type and/or skill type exceeds a threshold value, a response or action corresponding to that intent and/or skill type may be performed by the conversational entity.” Here, Suwandy teaches performing an action using a conversational entity based on an intent being identified. Further See Suwandy paragraph [0027] describing, “In some examples, a nearest neighbor score value may be calculated for each of the one or more intent types and/or skill types based on application of the KNN model. In some examples, if a nearest neighbor score value for an intent type and/or skill type exceeds a threshold value, a response or action corresponding to that intent and/or skill type may be performed by the conversational entity.” Here, Suwandy teaches performing an action or response using a conversational entity based on an intent being identified. Which can be seen as creating an intent recommendation.
Claim(s) 5-6 and 16-17 are rejected under 35 U.S.C. 103 as being unpatentable over Suwandy T. et al, in view of Cohen W. et al l, further in view of D'Alessandro A. et al, and further in view of Freed R. et al, (US. Patent 11,756,553 B2) filed on September 7, 2020.
Claim 5:
Regarding claim 5, Suwandy in view of Cohen and D’Alessandro teaches the limitations in claim 3.
Neither Suwandy, Cohen or D’Alessandro appears to teach the further limitations of this claim. However, Freed in the same field of art teaches “method of claim 3, wherein the comparing operation further comprises performing a check to determine is there exists an existing item with a similarity(intent_embeddingsi, item_embeddingj) that is greater than the given threshold.”
See Freed in column 4 lines 35-38 where it describes “Data enhancement module 110 may form a new intent by determining the similarity score among two or more existing intents exceeds a pre-defined threshold. Data enhancement module 110 may keep the root verb as a new intent if the similarity score among two or more existing intents exceeds the pre-defined threshold” Here, Freed shows checking to see if two pre-existing intents have a similarity that exceeds a given threshold. Using the broadest reasonable interpretation the similarity((intent_embeddingsi, item_embeddingj) is being interpreted as similarity between two intent embeddings, the item embedding being a previous item of the matrix. In an analogous system the intents of Freed can be from intent embeddings described in the previous claims.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the base reference of Suwandy with the teachings of Cohen, D’Alessandro and Freed by using Suwandy’s teachings of converting an intent into and embedding, comparing the embedding and training a model, and incorporate with Cohen and D’Alessandro’s teachings of mapping items to a similar and incrementing a corresponding count of the similar, and using a second trained model to perform inferencing, and further Freed’s teachings of performing a check for existing similarity between existing items.
One of ordinary skill in the art would be motivated to do so because by integrating Freed’s frameworks into the methods of Suwandy, Cohen, and D’Alessandro, one with ordinary skill in the art would bring “performing an evidence-based analysis of a chatbot's existing training data and identifying opportunities for improvement by identifying intents that can be refactored into intents with entities.” (Freed, column 1 lines 66-67 and column 2 lines 1-3), “taking utterances that a user intends to route differently in the chatbot and optimizing the use of intents and entities to do that routing.” (Freed, column 2 lines 4-6), “maintaining and improving a testing score (e.g., a k-fold score) of a training data set with the intention of increasing a blind score by being more generalizable and less confusable” (Freed, column 2 lines 7-10), and “improving training data by moving from more intents/few entities to less intents/more entities.” (Freed, column 2 lines 11-12).
Claim 6:
Regarding claim 6, Suwandy in view of Cohen and D’Alessandro teaches the limitations in claim 3.
Neither Suwandy, Cohen or D’Alessandro appears to teach the further limitations of this claim. However, Freed in the same field of art teaches “method of claim 3, further comprising repeating the encoding, comparing, mapping, creating the new item, and creating the matrix operations, for each workspace to update the matrix with each pair of training workspace and item”
See Freed column 5 lines 47-67 and column 6 lines 1-9 where it describes “In one or more embodiments, data enhancement module 110 is configured to generate a set of new training data based on the set of new intents and entities for chatbot server 104. Data enhancement module 110 may perform a test (e.g., a k-fold test or k-fold cross-validation) of the set of training data and the set of new training data. For example, k-fold cross-validation may be where a given data set is split into a K number of sections/folds where each fold is used as a testing set at some point. In the first iteration, the first fold may be used to test the model and the rest may be used to train the model. In the second iteration, 2nd fold is used as the testing set while the rest serve as the training set. This process is repeated until each fold has been used as the testing set. Data enhancement module 110 may analyze the tested results. Data enhancement module 110 may compare the testing results. Data enhancement module 110 may test the new training data against a k-fold testing algorithm to determine a validity score (needed to map to a new intent) of the new training data. Data enhancement module 110 may perform an action based on the validity score. Data enhancement module 110 may compare the test results by noting that an utterance previously resolved to a previous intent now needs to be resolved to a new intent. Data enhancement module 110 may provide a manual review indication for an utterance that does not include any entity listed. Data enhancement module 110 may automatically update the training for chatbot server 104 with the new training data when data enhancement module 110 determines the accuracy based on the testing of the new training data is maintained or improved.” Here, Freed teaches a process of repeating taking in new inputs, which in an analogous system could be encoding a user intent, comparing, testing new training data to map to an intent, resolving to a new intent which can be seen as creating a new item, and does the creating matrix operations described before. The repeat of these operations update the training data(s) of the data enhancement module to then update the chatbot server, which can be a matrix as known in the art.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the base reference of Suwandy with the teachings of Cohen, D’Alessandro and Freed by using Suwandy’s teachings of converting an intent into and embedding, comparing the embedding and training a model, and incorporate with Cohen and D’Alessandro’s teachings of mapping items to a similar and incrementing a corresponding count of the similar, and using a second trained model to perform inferencing, and further Freed’s teachings of repeating operations to update workspaces of a matrix.
One of ordinary skill in the art would be motivated to do so because by integrating Freed’s frameworks into the methods of Suwandy, Cohen, and D’Alessandro, one with ordinary skill in the art would bring “performing an evidence-based analysis of a chatbot's existing training data and identifying opportunities for improvement by identifying intents that can be refactored into intents with entities.” (Freed, column 1 lines 66-67 and column 2 lines 1-3), “taking utterances that a user intends to route differently in the chatbot and optimizing the use of intents and entities to do that routing.” (Freed, column 2 lines 4-6), “maintaining and improving a testing score (e.g., a k-fold score) of a training data set with the intention of increasing a blind score by being more generalizable and less confusable” (Freed, column 2 lines 7-10), and “improving training data by moving from more intents/few entities to less intents/more entities.” (Freed, column 2 lines 11-12).
Claim 16:
Regarding claim 16, Suwandy in view of Cohen and D’Alessandro teaches the limitations in claim 14.
Neither Suwandy, Cohen or D’Alessandro appears to teach the further limitations of this claim. However, Freed in the same field of art teaches “The system of claim 14, wherein the comparing operation further comprises performing a check to determine is there exists an existing item with a similarity(intent_embeddingsi, item_embeddingj) that is greater than the given threshold.”
See Freed in column 4 lines 35-38 where it describes “Data enhancement module 110 may form a new intent by determining the similarity score among two or more existing intents exceeds a pre-defined threshold. Data enhancement module 110 may keep the root verb as a new intent if the similarity score among two or more existing intents exceeds the pre-defined threshold” Here, Freed shows checking to see if two pre-existing intents have a similarity that exceeds a given threshold. Using the broadest reasonable interpretation the similarity((intent_embeddingsi, item_embeddingj) is being interpreted as similarity between two intent embeddings, the item embedding being a previous item of the matrix. In an analogous system the intents of Freed can be from intent embeddings described in the previous claims.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the base reference of Suwandy with the teachings of Cohen, D’Alessandro and Freed by using Suwandy’s teachings of converting an intent into and embedding, comparing the embedding and training a model, and incorporate with Cohen and D’Alessandro’s teachings of mapping items to a similar and incrementing a corresponding count of the similar, and using a second trained model to perform inferencing, and further Freed’s teachings of performing a check for existing similarity between existing items.
One of ordinary skill in the art would be motivated to do so because by integrating Freed’s frameworks into the methods of Suwandy, Cohen, and D’Alessandro, one with ordinary skill in the art would bring “performing an evidence-based analysis of a chatbot's existing training data and identifying opportunities for improvement by identifying intents that can be refactored into intents with entities.” (Freed, column 1 lines 66-67 and column 2 lines 1-3), “taking utterances that a user intends to route differently in the chatbot and optimizing the use of intents and entities to do that routing.” (Freed, column 2 lines 4-6), “maintaining and improving a testing score (e.g., a k-fold score) of a training data set with the intention of increasing a blind score by being more generalizable and less confusable” (Freed, column 2 lines 7-10), and “improving training data by moving from more intents/few entities to less intents/more entities.” (Freed, column 2 lines 11-12).
Claim 17:
Regarding claim 17, Suwandy in view of Cohen and D’Alessandro teaches the limitations in claim 14.
Neither Suwandy, Cohen or D’Alessandro appears to teach the further limitations of this claim. However, Freed in the same field of art teaches “The system of claim 14, the operations further comprising repeating the encoding, comparing, mapping, creating the new item, and creating the matrix operations, for each workspace to update the matrix with each pair of training workspace and item”
See Freed column 5 lines 47-67 and column 6 lines 1-9 where it describes “In one or more embodiments, data enhancement module 110 is configured to generate a set of new training data based on the set of new intents and entities for chatbot server 104. Data enhancement module 110 may perform a test (e.g., a k-fold test or k-fold cross-validation) of the set of training data and the set of new training data. For example, k-fold cross-validation may be where a given data set is split into a K number of sections/folds where each fold is used as a testing set at some point. In the first iteration, the first fold may be used to test the model and the rest may be used to train the model. In the second iteration, 2nd fold is used as the testing set while the rest serve as the training set. This process is repeated until each fold has been used as the testing set. Data enhancement module 110 may analyze the tested results. Data enhancement module 110 may compare the testing results. Data enhancement module 110 may test the new training data against a k-fold testing algorithm to determine a validity score (needed to map to a new intent) of the new training data. Data enhancement module 110 may perform an action based on the validity score. Data enhancement module 110 may compare the test results by noting that an utterance previously resolved to a previous intent now needs to be resolved to a new intent. Data enhancement module 110 may provide a manual review indication for an utterance that does not include any entity listed. Data enhancement module 110 may automatically update the training for chatbot server 104 with the new training data when data enhancement module 110 determines the accuracy based on the testing of the new training data is maintained or improved.” Here, Freed teaches a process of repeating taking in new inputs, which in an analogous system could be encoding a user intent, comparing, testing new training data to map to an intent, resolving to a new intent which can be seen as creating a new item, and does the creating matrix operations described before. The repeat of these operations update the training data(s) of the data enhancement module to then update the chatbot server, which can be a matrix as known in the art.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the base reference of Suwandy with the teachings of Cohen, D’Alessandro and Freed by using Suwandy’s teachings of converting an intent into and embedding, comparing the embedding and training a model, and incorporate with Cohen and D’Alessandro’s teachings of mapping items to a similar and incrementing a corresponding count of the similar, and using a second trained model to perform inferencing, and further Freed’s teachings of repeating operations to update workspaces of a matrix.
One of ordinary skill in the art would be motivated to do so because by integrating Freed’s frameworks into the methods of Suwandy, Cohen, and D’Alessandro, one with ordinary skill in the art would bring “performing an evidence-based analysis of a chatbot's existing training data and identifying opportunities for improvement by identifying intents that can be refactored into intents with entities.” (Freed, column 1 lines 66-67 and column 2 lines 1-3), “taking utterances that a user intends to route differently in the chatbot and optimizing the use of intents and entities to do that routing.” (Freed, column 2 lines 4-6), “maintaining and improving a testing score (e.g., a k-fold score) of a training data set with the intention of increasing a blind score by being more generalizable and less confusable” (Freed, column 2 lines 7-10), and “improving training data by moving from more intents/few entities to less intents/more entities.” (Freed, column 2 lines 11-12).
Claim(s) 7-9 and 18-20 are rejected under 35 U.S.C. 103 as being unpatentable over Suwandy T. et al, in view of Cohen W. et al l, further in view of D'Alessandro A. et al, and further in view of Kumar D. et al, (US. Patent Application Publication 20240419901 A1) filed on June 15, 2023.
Claim 7:
Regarding claim 7, Suwandy in view of Cohen and D’Alessandro teaches the limitations in claim 3.
Further, Suwandy teaches “method of claim 3, further comprising: mapping a given client workspace into the matrix”
See Suwandy in paragraph [0027] describing “During operation, the conversational computing service may process incoming natural language user inputs that have been received via conversational entities. When a new natural language input is received, that input may be encoded as an embedding. In some examples, the embedding may be compressed into a binary form. The embedding may then be added to the embedding library.” Here, Suwandy maps a client workspace into a matrix by taking in user inputs which can be seen as a client workspace, converting it into an embedding and then adding it to an embedding library, which previously is stated to already be perceived as a matrix, which can be seen as mapping into a matrix.
Further, Suwandy teaches, “conducting a search of the training workspace that contains a similar set of items compared to the given client workspace”
See Suwandy in paragraph [0027] where it describes “During operation, the conversational computing service may process incoming natural language user inputs that have been received via conversational entities. When a new natural language input is received, that input may be encoded as an embedding. In some examples, the embedding may be compressed into a binary form. The embedding may then be added to the embedding library. Once added to the embedding library, a similarity score model may be applied to the new embedding and one or more embeddings from the examples that were provided by the conversational entity developer. In some examples, the similarity score model may be a cosine model. In other examples, the similarity score model may be a Hamming model. A similarity score may be calculated between the new embedding and one or more of the example embeddings. A K Nearest Neighbors (KNN) model may then be utilized to identify one or more intent types and/or skill types that correspond to the new embedding. In some examples, a nearest neighbor score value may be calculated for each of the one or more intent types and/or skill types based on application of the KNN model.” Here, Suwandy shows using a K Nearest Neighbors model, which is used to perform searches as understood in the art, to look for intents that correspond to the new embedding which in this instance is the client workspace as mentioned. The search for similar items is done using similarity scores and nearest neighbor score.
Neither Suwandy, Cohen or D’Alessandro appears to teach the “grouping utterances of client log data into clusters and mapping the clusters into items of the similar training workspaces using the corresponding embeddings; recommending the first and second intents corresponding to the mapped items that exist in the similar training workspace and are absent from the client workspace.”
However, Kumar in the same field of art teaches, “grouping utterances of client log data into clusters and mapping the clusters into items of the similar training workspaces using the corresponding embeddings”
See Kumar in paragraph [0017] describing, “In some implementations, when preprocessing the text data with the one or more preprocessing techniques to generate the key intents, the transformation system may identify utterances in the text data, and may generate parts of speech tags for the text data (e.g., by assigning parts of speech tags that correspond to particular parts of speech, such as nouns, verbs, adjectives, adverbs, and/or the like).” Here, Kumar establishes utterances being a part of the text data which can be used to generate intents, in these, these utterances are grouped. Further, see Kumar in paragraph [0018] describing, “the transformation system 110 may convert the preprocessed data and the key intents into embeddings.” The intents mentioned can be turned into embeddings. Further, see Kumar in Figure 2,
PNG
media_image1.png
397
491
media_image1.png
Greyscale
. In this figure given by Kumar, we can see clusters of sets of observations which as mentioned before can be seen as training workspaces. The new observation, which also as mentioned can be seen as a client workspace consists of intents which come from utterances being mapped to other observations.
Further, Kumar teaches, “recommending the first and second intents corresponding to the mapped items that exist in the similar training workspace and are absent from the client workspace.”
See Kumar in paragraph [0041], where it describes “In some implementations, the trained machine learning model 225 may classify (e.g., cluster) the new observation in a cluster, as shown by reference number 240. The observations within a cluster may have a threshold degree of similarity. As an example, if the machine learning system classifies the new observation in a first cluster (e.g., a preprocessed data cluster), then the machine learning system may provide a first recommendation.” Further, see Kumar in paragraph [0042] describing, “As another example, if the machine learning system were to classify the new observation in a second cluster (e.g., a key intents cluster), then the machine learning system may provide a second (e.g., different) recommendation” Here, for the new observation, as mentioned before being seen as a client workspace, if the recommendation is made using clusters instead of the new observation in itself it can be seen that the recommendation exist in a similar training workspace and is absent from the client workspace. Kumar here also teaches a first and second recommendation being made off intents.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the base reference of Suwandy with the teachings of Cohen, D’Alessandro and Kumar by using Suwandy’s teachings of converting an intent into and embedding, comparing the embedding and training a model, and incorporate with Cohen and D’Alessandro’s teachings of mapping items to a similar and incrementing a corresponding count of the similar, and using a second trained model to perform inferencing, and further Kumar’s teachings of grouping utterances into clusters and mapping clusters into training workspaces, and recommending first and second intents.
One of ordinary skill in the art would be motivated to do so because by integrating Kumar’s frameworks into the methods of Suwandy, Cohen, and D’Alessandro, one with ordinary skill in the art would bring a “transformation system [that] generates a conversation summary from text data using a language transformation model” (Kumar, paragraph [0011]), “transformation system [that] may process the preprocessed data and the key intents, with a language model, to generate a conversation summary for the transcript. The transformation system may utilize the conversation summary to understand a customer journey, a customer issue, and/or a customer need, to provide an improved search experience, to identify novel categories, and/or the like” (Kumar, paragraph [0011]), and a “transformation system [that] may conserve computing resources, networking resources, and/or other resources that would have otherwise been consumed by failing to identify meaningful information and/or phrases in text data, failing to identify summaries of conversations provided in text data, being unable to utilize the meaningful information and/or phrases and the summaries of conversations, and/or the like.” (Kumar, paragraph [0011]).
Claim 8:
Regarding claim 8, Suwandy in view of Cohen and D’Alessandro teaches the limitations in claim 3.
Further, Suwandy teaches “method of claim 3, further comprising: mapping a given client workspace into the matrix”
See Suwandy in paragraph [0027] describing “During operation, the conversational computing service may process incoming natural language user inputs that have been received via conversational entities. When a new natural language input is received, that input may be encoded as an embedding. In some examples, the embedding may be compressed into a binary form. The embedding may then be added to the embedding library.” Here, Suwandy maps a client workspace into a matrix by taking in user inputs which can be seen as a client workspace, converting it into an embedding and then adding it to an embedding library, which previously is stated to already be perceived as a matrix, which can be seen as mapping into a matrix.
Further, Suwandy teaches, “conducting a search of the training workspace(s) that contain a similar set of items in the matrix compared to the given client workspace”
See Suwandy in paragraph [0027] where it describes “During operation, the conversational computing service may process incoming natural language user inputs that have been received via conversational entities. When a new natural language input is received, that input may be encoded as an embedding. In some examples, the embedding may be compressed into a binary form. The embedding may then be added to the embedding library. Once added to the embedding library, a similarity score model may be applied to the new embedding and one or more embeddings from the examples that were provided by the conversational entity developer. In some examples, the similarity score model may be a cosine model. In other examples, the similarity score model may be a Hamming model. A similarity score may be calculated between the new embedding and one or more of the example embeddings. A K Nearest Neighbors (KNN) model may then be utilized to identify one or more intent types and/or skill types that correspond to the new embedding. In some examples, a nearest neighbor score value may be calculated for each of the one or more intent types and/or skill types based on application of the KNN model.” Here, Suwandy shows using a K Nearest Neighbors model, which is used to perform searches as understood in the art, to look for intents that correspond to the new embedding which in this instance is the client workspace as mentioned. The search for similar items is done using similarity scores and nearest neighbor score.
Neither Suwandy, Cohen or D’Alessandro appears to teach the “recommending the first and second intents corresponding to the mapped items that exist in the similar training workspace and are absent from the client workspace”.
However, Kumar in the same field of art teaches, “recommending the first and second intents corresponding to the mapped items that exist in the similar training workspace and are absent from the client workspace”
See Kumar in paragraph [0017] describing, “In some implementations, when preprocessing the text data with the one or more preprocessing techniques to generate the key intents, the transformation system may identify utterances in the text data, and may generate parts of speech tags for the text data (e.g., by assigning parts of speech tags that correspond to particular parts of speech, such as nouns, verbs, adjectives, adverbs, and/or the like).” Here, Kumar establishes utterances being a part of the text data which can be used to generate intents, in these, these utterances are grouped. Further, see Kumar in paragraph [0018] describing, “the transformation system 110 may convert the preprocessed data and the key intents into embeddings.” The intents mentioned can be turned into embeddings. Further, see Kumar in Figure 2,
PNG
media_image1.png
397
491
media_image1.png
Greyscale
. In this figure given by Kumar, we can see clusters of sets of observations which as mentioned before can be seen as training workspaces. The new observation, which also as mentioned can be seen as a client workspace consists of intents which come from utterances being mapped to other observations.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the base reference of Suwandy with the teachings of Cohen, D’Alessandro and Kumar by using Suwandy’s teachings of converting an intent into and embedding, comparing the embedding and training a model, and incorporate with Cohen and D’Alessandro’s teachings of mapping items to a similar and incrementing a corresponding count of the similar, and using a second trained model to perform inferencing, and further Kumar’s teaching of recommending first and second intents.
One of ordinary skill in the art would be motivated to do so because by integrating Kumar’s frameworks into the methods of Suwandy, Cohen, and D’Alessandro, one with ordinary skill in the art would bring a “transformation system [that] generates a conversation summary from text data using a language transformation model” (Kumar, paragraph [0011]), “transformation system [that] may process the preprocessed data and the key intents, with a language model, to generate a conversation summary for the transcript. The transformation system may utilize the conversation summary to understand a customer journey, a customer issue, and/or a customer need, to provide an improved search experience, to identify novel categories, and/or the like” (Kumar, paragraph [0011]), and a “transformation system [that] may conserve computing resources, networking resources, and/or other resources that would have otherwise been consumed by failing to identify meaningful information and/or phrases in text data, failing to identify summaries of conversations provided in text data, being unable to utilize the meaningful information and/or phrases and the summaries of conversations, and/or the like.” (Kumar, paragraph [0011]).
Claim 9:
Regarding claim 9, Suwandy in view of Cohen and D’Alessandro teaches the limitations in claim 3.
Neither Suwandy, Cohen or D’Alessandro appears to teach the further limitations of this claim. However, Kumar in the same field of art teaches, “method of claim 3, further comprising: grouping utterances of client log data into clusters and mapping the clusters into items of the training workspace using the corresponding embeddings”
See Kumar in paragraph [0017] describing, “In some implementations, when preprocessing the text data with the one or more preprocessing techniques to generate the key intents, the transformation system may identify utterances in the text data, and may generate parts of speech tags for the text data (e.g., by assigning parts of speech tags that correspond to particular parts of speech, such as nouns, verbs, adjectives, adverbs, and/or the like).” Here, Kumar establishes utterances being a part of the text data which can be used to generate intents, in these, these utterances are grouped. Further, see Suwandy in paragraph [0018] describing, “the transformation system 110 may convert the preprocessed data and the key intents into embeddings.” The intents mentioned can be turned into embeddings. Further, see Kumar in Figure 2,
PNG
media_image1.png
397
491
media_image1.png
Greyscale
. In this figure given by Kumar, we can see clusters of sets of observations which as mentioned before can be seen as training workspaces. The new observation, which also as mentioned can be seen as a client workspace consists of intents which come from utterances being mapped to other observations.
Further, Kumar teaches, “recommending the first and second intents corresponding to the mapped items that exist in the training workspace.”
See Kumar in paragraph [0041], where it describes “In some implementations, the trained machine learning model 225 may classify (e.g., cluster) the new observation in a cluster, as shown by reference number 240. The observations within a cluster may have a threshold degree of similarity. As an example, if the machine learning system classifies the new observation in a first cluster (e.g., a preprocessed data cluster), then the machine learning system may provide a first recommendation.” Further, see Kumar in paragraph [0042] describing, “As another example, if the machine learning system were to classify the new observation in a second cluster (e.g., a key intents cluster), then the machine learning system may provide a second (e.g., different) recommendation” Here, for the new observation, as mentioned before being seen as a client workspace, if the recommendation is made using clusters instead of the new observation in itself it can be seen that the recommendation exist in a similar training workspace. Suwandy here also teaches a first and second recommendation being made off intents.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the base reference of Suwandy with the teachings of Cohen, D’Alessandro and Kumar by using Suwandy’s teachings of converting an intent into and embedding, comparing the embedding and training a model, and incorporate with Cohen and D’Alessandro’s teachings of mapping items to a similar and incrementing a corresponding count of the similar, and using a second trained model to perform inferencing, and further Kumar’s teachings of grouping utterances into clusters and mapping clusters into training workspaces, and recommending first and second intents.
One of ordinary skill in the art would be motivated to do so because by integrating Kumar’s frameworks into the methods of Suwandy, Cohen, and D’Alessandro, one with ordinary skill in the art would bring a “transformation system [that] generates a conversation summary from text data using a language transformation model” (Kumar, paragraph [0011]), “transformation system [that] may process the preprocessed data and the key intents, with a language model, to generate a conversation summary for the transcript. The transformation system may utilize the conversation summary to understand a customer journey, a customer issue, and/or a customer need, to provide an improved search experience, to identify novel categories, and/or the like” (Kumar, paragraph [0011]), and a “transformation system [that] may conserve computing resources, networking resources, and/or other resources that would have otherwise been consumed by failing to identify meaningful information and/or phrases in text data, failing to identify summaries of conversations provided in text data, being unable to utilize the meaningful information and/or phrases and the summaries of conversations, and/or the like.” (Kumar, paragraph [0011]).
Claim 18:
Regarding claim 18, Suwandy in view of Cohen and D’Alessandro teaches the limitations in claim 14.
Further, Suwandy teaches “The system of claim 14, the operations further comprising: mapping a given client workspace into the matrix”
See Suwandy in paragraph [0027] describing “During operation, the conversational computing service may process incoming natural language user inputs that have been received via conversational entities. When a new natural language input is received, that input may be encoded as an embedding. In some examples, the embedding may be compressed into a binary form. The embedding may then be added to the embedding library.” Here, Suwandy maps a client workspace into a matrix by taking in user inputs which can be seen as a client workspace, converting it into an embedding and then adding it to an embedding library, which previously is stated to already be perceived as a matrix, which can be seen as mapping into a matrix.
Further, Suwandy teaches, “conducting a search of the training workspace that contains a similar set of items compared to the given client workspace”
See Suwandy in paragraph [0027] where it describes “During operation, the conversational computing service may process incoming natural language user inputs that have been received via conversational entities. When a new natural language input is received, that input may be encoded as an embedding. In some examples, the embedding may be compressed into a binary form. The embedding may then be added to the embedding library. Once added to the embedding library, a similarity score model may be applied to the new embedding and one or more embeddings from the examples that were provided by the conversational entity developer. In some examples, the similarity score model may be a cosine model. In other examples, the similarity score model may be a Hamming model. A similarity score may be calculated between the new embedding and one or more of the example embeddings. A K Nearest Neighbors (KNN) model may then be utilized to identify one or more intent types and/or skill types that correspond to the new embedding. In some examples, a nearest neighbor score value may be calculated for each of the one or more intent types and/or skill types based on application of the KNN model.” Here, Suwandy shows using a K Nearest Neighbors model, which is used to perform searches as understood in the art, to look for intents that correspond to the new embedding which in this instance is the client workspace as mentioned. The search for similar items is done using similarity scores and nearest neighbor score.
Neither Suwandy, Cohen or D’Alessandro appears to teach “grouping utterances of client log data into clusters and mapping the clusters into items of the similar training workspaces using the corresponding embeddings; recommending the first and second intents corresponding to the mapped items that exist in the similar training workspace and are absent from the client workspace”.
However, Kumar in the same field of art teaches, “grouping utterances of client log data into clusters and mapping the clusters into items of the similar training workspaces using the corresponding embeddings”
See Kumar in paragraph [0017] describing, “In some implementations, when preprocessing the text data with the one or more preprocessing techniques to generate the key intents, the transformation system may identify utterances in the text data, and may generate parts of speech tags for the text data (e.g., by assigning parts of speech tags that correspond to particular parts of speech, such as nouns, verbs, adjectives, adverbs, and/or the like).” Here, Kumar establishes utterances being a part of the text data which can be used to generate intents, in these, these utterances are grouped. Further, see Kumar in paragraph [0018] describing, “the transformation system 110 may convert the preprocessed data and the key intents into embeddings.” The intents mentioned can be turned into embeddings. Further, see Kumar in Figure 2,
PNG
media_image1.png
397
491
media_image1.png
Greyscale
. In this figure given by Kumar, we can see clusters of sets of observations which as mentioned before can be seen as training workspaces. The new observation, which also as mentioned can be seen as a client workspace consists of intents which come from utterances being mapped to other observations.
Further, Kumar teaches, “recommending the first and second intents corresponding to the mapped items that exist in the similar training workspace and are absent from the client workspace.”
See Kumar in paragraph [0041], where it describes “In some implementations, the trained machine learning model 225 may classify (e.g., cluster) the new observation in a cluster, as shown by reference number 240. The observations within a cluster may have a threshold degree of similarity. As an example, if the machine learning system classifies the new observation in a first cluster (e.g., a preprocessed data cluster), then the machine learning system may provide a first recommendation.” Further, see Kumar in paragraph [0042] describing, “As another example, if the machine learning system were to classify the new observation in a second cluster (e.g., a key intents cluster), then the machine learning system may provide a second (e.g., different) recommendation” Here, for the new observation, as mentioned before being seen as a client workspace, if the recommendation is made using clusters instead of the new observation in itself it can be seen that the recommendation exist in a similar training workspace and is absent from the client workspace. Kumar here also teaches a first and second recommendation being made off intents.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the base reference of Suwandy with the teachings of Cohen, D’Alessandro and Kumar by using Suwandy’s teachings of converting an intent into and embedding, comparing the embedding and training a model, and incorporate with Cohen and D’Alessandro’s teachings of mapping items to a similar and incrementing a corresponding count of the similar, and using a second trained model to perform inferencing, and further Kumar’s teachings of grouping utterances into clusters and mapping clusters into training workspaces, and recommending first and second intents.
One of ordinary skill in the art would be motivated to do so because by integrating Kumar’s frameworks into the methods of Suwandy, Cohen, and D’Alessandro, one with ordinary skill in the art would bring a “transformation system [that] generates a conversation summary from text data using a language transformation model” (Kumar, paragraph [0011]), “transformation system [that] may process the preprocessed data and the key intents, with a language model, to generate a conversation summary for the transcript. The transformation system may utilize the conversation summary to understand a customer journey, a customer issue, and/or a customer need, to provide an improved search experience, to identify novel categories, and/or the like” (Kumar, paragraph [0011]), and a “transformation system [that] may conserve computing resources, networking resources, and/or other resources that would have otherwise been consumed by failing to identify meaningful information and/or phrases in text data, failing to identify summaries of conversations provided in text data, being unable to utilize the meaningful information and/or phrases and the summaries of conversations, and/or the like.” (Kumar, paragraph [0011]).
Claim 19:
Regarding claim 19, Suwandy in view of Cohen and D’Alessandro teaches the limitations in claim 3.
Further, Suwandy teaches “The system of claim 14, the operations further comprising: mapping a given client workspace into the matrix”
See Suwandy in paragraph [0027] describing “During operation, the conversational computing service may process incoming natural language user inputs that have been received via conversational entities. When a new natural language input is received, that input may be encoded as an embedding. In some examples, the embedding may be compressed into a binary form. The embedding may then be added to the embedding library.” Here, Suwandy maps a client workspace into a matrix by taking in user inputs which can be seen as a client workspace, converting it into an embedding and then adding it to an embedding library, which previously is stated to already be perceived as a matrix, which can be seen as mapping into a matrix.
Further, Suwandy teaches, “conducting a search of the training workspace(s) that contain a similar set of items in the matrix compared to the given client workspace”
See Suwandy in paragraph [0027] where it describes “During operation, the conversational computing service may process incoming natural language user inputs that have been received via conversational entities. When a new natural language input is received, that input may be encoded as an embedding. In some examples, the embedding may be compressed into a binary form. The embedding may then be added to the embedding library. Once added to the embedding library, a similarity score model may be applied to the new embedding and one or more embeddings from the examples that were provided by the conversational entity developer. In some examples, the similarity score model may be a cosine model. In other examples, the similarity score model may be a Hamming model. A similarity score may be calculated between the new embedding and one or more of the example embeddings. A K Nearest Neighbors (KNN) model may then be utilized to identify one or more intent types and/or skill types that correspond to the new embedding. In some examples, a nearest neighbor score value may be calculated for each of the one or more intent types and/or skill types based on application of the KNN model.” Here, Suwandy shows using a K Nearest Neighbors model, which is used to perform searches as understood in the art, to look for intents that correspond to the new embedding which in this instance is the client workspace as mentioned. The search for similar items is done using similarity scores and nearest neighbor score.
Neither Suwandy, Cohen or D’Alessandro appears to teach “recommending the first and second intents corresponding to the mapped items that exist in the similar training workspace and are absent from the client workspace”.
However, Kumar in the same field of art teaches, “recommending the first and second intents corresponding to the mapped items that exist in the similar training workspace and are absent from the client workspace”
See Kumar in paragraph [0017] describing, “In some implementations, when preprocessing the text data with the one or more preprocessing techniques to generate the key intents, the transformation system may identify utterances in the text data, and may generate parts of speech tags for the text data (e.g., by assigning parts of speech tags that correspond to particular parts of speech, such as nouns, verbs, adjectives, adverbs, and/or the like).” Here, Kumar establishes utterances being a part of the text data which can be used to generate intents, in these, these utterances are grouped. Further, see Kumar in paragraph [0018] describing, “the transformation system 110 may convert the preprocessed data and the key intents into embeddings.” The intents mentioned can be turned into embeddings. Further, see Kumar in Figure 2,
PNG
media_image1.png
397
491
media_image1.png
Greyscale
. In this figure given by Kumar, we can see clusters of sets of observations which as mentioned before can be seen as training workspaces. The ne observation, which also as mentioned can be seen as a client workspace consists of intents which come from utterances being mapped to other observations.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the base reference of Suwandy with the teachings of Cohen, D’Alessandro and Kumar by using Suwandy’s teachings of converting an intent into and embedding, comparing the embedding and training a model, and incorporate with Cohen and D’Alessandro’s teachings of mapping items to a similar and incrementing a corresponding count of the similar, and using a second trained model to perform inferencing, and further Kumar’s teaching of recommending first and second intents.
One of ordinary skill in the art would be motivated to do so because by integrating Kumar’s frameworks into the methods of Suwandy, Cohen, and D’Alessandro, one with ordinary skill in the art would bring a “transformation system [that] generates a conversation summary from text data using a language transformation model” (Kumar, paragraph [0011]), “transformation system [that] may process the preprocessed data and the key intents, with a language model, to generate a conversation summary for the transcript. The transformation system may utilize the conversation summary to understand a customer journey, a customer issue, and/or a customer need, to provide an improved search experience, to identify novel categories, and/or the like” (Kumar, paragraph [0011]), and a “transformation system [that] may conserve computing resources, networking resources, and/or other resources that would have otherwise been consumed by failing to identify meaningful information and/or phrases in text data, failing to identify summaries of conversations provided in text data, being unable to utilize the meaningful information and/or phrases and the summaries of conversations, and/or the like.” (Kumar, paragraph [0011]).
Claim 20:
Regarding claim 20, Suwandy in view of Cohen and D’Alessandro teaches the limitations in claim 14.
Neither Suwandy, Cohen or D’Alessandro appears to teach the further limitations of this claim. However, Kumar in the same field of art teaches, “The system of claim 14, the operations further comprising: grouping utterances of client log data into clusters and mapping the clusters into items of the training workspace using the corresponding embeddings”
See Kumar in paragraph [0017] describing, “In some implementations, when preprocessing the text data with the one or more preprocessing techniques to generate the key intents, the transformation system may identify utterances in the text data, and may generate parts of speech tags for the text data (e.g., by assigning parts of speech tags that correspond to particular parts of speech, such as nouns, verbs, adjectives, adverbs, and/or the like).” Here, Kumar establishes utterances being a part of the text data which can be used to generate intents, in these, these utterances are grouped. Further, see Suwandy in paragraph [0018] describing, “the transformation system 110 may convert the preprocessed data and the key intents into embeddings.” The intents mentioned can be turned into embeddings. Further, see Kumar in Figure 2,
PNG
media_image1.png
397
491
media_image1.png
Greyscale
. In this figure given by Kumar, we can see clusters of sets of observations which as mentioned before can be seen as training workspaces. The new observation, which also as mentioned can be seen as a client workspace consists of intents which come from utterances being mapped to other observations.
Further, Kumar teaches, “recommending the first and second intents corresponding to the mapped items that exist in the training workspace.”
See Kumar in paragraph [0041], where it describes “In some implementations, the trained machine learning model 225 may classify (e.g., cluster) the new observation in a cluster, as shown by reference number 240. The observations within a cluster may have a threshold degree of similarity. As an example, if the machine learning system classifies the new observation in a first cluster (e.g., a preprocessed data cluster), then the machine learning system may provide a first recommendation.” Further, see Kumar in paragraph [0042] describing, “As another example, if the machine learning system were to classify the new observation in a second cluster (e.g., a key intents cluster), then the machine learning system may provide a second (e.g., different) recommendation” Here, for the new observation, as mentioned before being seen as a client workspace, if the recommendation is made using clusters instead of the new observation in itself it can be seen that the recommendation exist in a similar training workspace. Suwandy here also teaches a first and second recommendation being made off intents.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the base reference of Suwandy with the teachings of Cohen, D’Alessandro and Kumar by using Suwandy’s teachings of converting an intent into and embedding, comparing the embedding and training a model, and incorporate with Cohen and D’Alessandro’s teachings of mapping items to a similar and incrementing a corresponding count of the similar, and using a second trained model to perform inferencing, and further Kumar’s teachings of grouping utterances into clusters and mapping clusters into training workspaces, and recommending first and second intents.
One of ordinary skill in the art would be motivated to do so because by integrating Kumar’s frameworks into the methods of Suwandy, Cohen, and D’Alessandro, one with ordinary skill in the art would bring a “transformation system [that] generates a conversation summary from text data using a language transformation model” (Kumar, paragraph [0011]), “transformation system [that] may process the preprocessed data and the key intents, with a language model, to generate a conversation summary for the transcript. The transformation system may utilize the conversation summary to understand a customer journey, a customer issue, and/or a customer need, to provide an improved search experience, to identify novel categories, and/or the like” (Kumar, paragraph [0011]), and a “transformation system [that] may conserve computing resources, networking resources, and/or other resources that would have otherwise been consumed by failing to identify meaningful information and/or phrases in text data, failing to identify summaries of conversations provided in text data, being unable to utilize the meaningful information and/or phrases and the summaries of conversations, and/or the like.” (Kumar, paragraph [0011]).
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to HASSAN R SESAY whose telephone number is (571)272-8493. The examiner can normally be reached Monday-Friday 8am-5pm.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Jeanette Parker can be reached at (571) 270-3647. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/HASSAN RAMADAN SESAY/
Examiner, Art Unit 2146 /USMAAN SAEED/Supervisory Patent Examiner, Art Unit 2146
Prosecution Insights