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 .
Claims 1 - 20 are pending and claims 1, 8 and 15 are independent claims.
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 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more. The independent claims 1, 8 and 15 recite “receiving… transcripts associated with calls or chats; filtering, …; predicting, …; filtering, …; generating, …; assigning, …; ranking, …; identifying, …; and performing …” as drafted cover an abstract idea of data analysis/retrieval and mental steps. More specifically, the “receiving, by a device, transcripts associated with calls or chats; filtering, by the device, customer utterances from the transcripts; predicting, by the device, intents for the customer utterances, wherein the intents are associated with corresponding probabilities; filtering, by the device, the intents to remove intents with probabilities below a threshold and to generate a filtered set of intents; generating, by the device, features from the customer utterances and the filtered set of intents using a natural language processing technique; assigning, by the device, respective weights to the filtered set of intents based on a comparison between the features and the customer utterances; ranking, by the device, the filtered set of intents based on the respective weights to identify a primary intent and a secondary intent of one of the calls or the chats; identifying, by the device, a primary intent reason for the primary intent and a secondary intent reason for the secondary intent based on historical data and an association model; and performing, by the device, one or more actions based on the primary intent reason or the secondary intent reason” which requires just data analysis / retrieval step and mental process. For instance, one can mentally receive transcripts associated with calls or chats apply a filtering, and then predict intents using probability and statistics approaches, then filter the intents further to remove those intents with probabilities below a certain threshold and then to generate a filtered set of intents. Then one might mentally create features from the customer utterances and the filtered set of intents. The “natural language processing technique” is just an additional element. One can mentally assign respective weights to the filtered set of intents based on a comparison between the features and the customer utterances, and rank the filtered set of intents based on the respective weights to identify a primary intent and a secondary intent of one of the calls or the chats. The “device” in these claims is just another additional element. One can mentally identify a primary intent reason for the primary intent and a secondary intent reason for the secondary intent based on historical data and the “association model” here is also just another additional element. One can mentally make a decision to perform one or more actions based on the primary intent reason or the secondary intent reason identified. The claimed invention is, therefore, directed to an abstract idea and a mental process without significantly more and thus, claims 1, 8 and 15 are rejected under 35 U.S.C. 101.
Similarly, the dependent claims 2-7, 9-14 and 16-20 recite similar claim language as in claims 1, 8 and 15.
Claims 2 and 16 recite “utilizing a model to generate the features from the customer utterances and the filtered set of intents,” which requires just a mental or statistical steps of generating the features from the customer utterances and the filtered set of intents. A “model” here is just another additional element. Thus, these claims 2 and 16 are directed to an abstract idea.
Claims 3 and 17 which recite “assigning greater weights to intents, of the filtered set of intents, associated with a top feature compared to respective weights assigned to intents, of the filtered set of intents, associated with other features,” which also requires just a mental step assigning greater weights to intents which are associated with a top feature. Thus, claims 3 and 17 are directed to an abstract idea.
Claim 4 which recites “utilizing a small language model to predict the intents for the customer utterances,” which also requires just a mental step of predicting the intents for the customer utterances. Thus, claim 4 is directed to an abstract idea.
Claim 5 which recites “utilizing the historical data to train the association model prior to identifying the primary intent reason for the primary intent and the secondary intent reason for the secondary intent,” which also requires just a mental step of utilizing the historical data to identifying the primary intent reason for the primary intent and the secondary intent reason for the secondary intent. The “training” and “association model” are just additional elements. Thus, claim 5 is directed to an abstract idea.
Claim 6 which recites “the association model generates a reason score for each intent of the filtered set of intents based on a frequency of each intent being associated with the primary intent in the historical data,” which also requires just a mental or statistical step of generating a reason score for each intent of the filtered set of intents based on a frequency of each intent being associated with the primary intent in the historical data. The “association model” is just an additional element. Thus, claim 6 is directed to an abstract idea.
Claim 7 which recites “removing agent utterances from the transcripts,” which also requires just a mental step of removing agent utterances from the transcripts. Thus, claim 7 is directed to an abstract idea.
Claims 9 and 18 which recite “receive a new transcript that includes new customer utterances associated with a new call or a new chat, and process the new transcript, with the association model, to identify a new primary intent and a new primary intent reason for the new primary intent,” which also requires just a mental step of receiving a new transcript that includes new customer utterances associated with a new call or a new chat, and processing the new transcript to identify a new primary intent and a new primary intent reason for the new primary intent. The “association model” is just an additional element. Thus, claims 9 and 18 are directed to an abstract idea.
Claims 10 and 19 which recite “aggregate the customer utterances into dynamic windows prior to predicting the intents for the customer utterances,” which also requires just a mental or mathematical/statistical step of aggregating the customer utterances into dynamic windows prior to predicting the intents for the customer utterances. Thus, claims 10 and 19 are directed to an abstract idea.
Claims 11 and 20 which recite “utilize a similarity analysis to compare the features with the customer utterances to determine similarity scores, and assign the respective weights to the filtered set of intents based on the similarity scores and predefined weighting factors,” which also requires just a mathematical step of utilizing a similarity analysis to compare the features with the customer utterances to determine similarity scores, and assign the respective weights to the filtered set of intents based on the similarity scores and predefined weighting factors. Thus, claims 11 and 20 are directed to an abstract idea.
Claim 12 which recites “utilize cosine similarity to determine the similarity scores based on degrees of similarity between the features and the customer utterances,” which also requires just a mathematical step of applying cosine similarity to determine the similarity scores based on degrees of similarity between the features and the customer utterances. This involves some mathematical formula to find the cosine similarity to determine the similarity scores based on degrees of similarity between the features and the customer utterances and this can be performed using a conventional/generic (general-purpose) computer or using a simple calculator. Thus, claim 12 is directed to an abstract idea.
Claim 13 which recites “generate n-gram features ranging from a minimum length to a maximum length based on the customer utterances and the filtered set of intents,” which also requires just a mathematical step of generating n-gram features ranging from a minimum length to a maximum length based on the customer utterances and the filtered set of intents. This involves some mathematical formula to generate n-gram features ranging from a minimum length to a maximum length based on the customer utterances and the filtered set of intents and this can be performed using a conventional/generic (general-purpose) computer or using a simple calculator. Thus, claim 13 is directed to an abstract idea.
Claim 14 which recites “provide the primary intent reason or the secondary intent reason for display, identify an originating factor of the one of the calls or the chats based on the primary intent reason or the secondary intent reason, address a customer issue based on the primary intent reason or the secondary intent reason, generate a business metric based on the primary intent reason or the secondary intent reason, or retrain the association model based on the primary intent reason or the secondary intent reason,” which also requires just a mathematical step of generating n-gram features ranging from a minimum length to a maximum length based on the customer utterances and the filtered set of intents. Thus, claim 14 is directed to an abstract idea.
Thus, claims 1-20 as drafted cover a mental process and abstract idea of data gathering/retrieval and analysis/processing steps, and they are mental processes directed to an abstract idea of implementing some mathematical formulae for data processing and data analysis using a conventional/generic (general-purpose) computer as well and thus, all the claims are directed to an abstract idea.
This judicial exception is not integrated into a practical application. In particular, claims 1, 8 and 15 recite additional element of “device”, “association model”, “processor” and “non-transitory computer-readable medium” as per the independent claims. Accordingly, these additional elements do not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea. The claim is directed to an abstract idea.
Thus, taken alone, the additional elements do not amount to significantly more than the above-identified judicial exception (the abstract idea). Looking at the limitations as an ordered combination adds nothing that is not already present when looking at the elements taken individually. There is no indication that the combination of elements improves the functioning of a computer or improves any other technology. Their collective functions merely provide conventional general purpose computer implementation. Claims 1-20, are therefore not drawn to patent eligible subject matter as they are directed to an abstract idea without significantly more. Thus, the claimed invention is directed to an abstract idea and a mental process without significantly more and thus, claims 1-20 are rejected under 35 U.S.C. 101.
The claims do not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to the integration of the abstract idea into a practical application, the additional element of using a computer is noted as a general computer as noted. Mere instructions to apply an exception using a generic computer component (Spec, para 65, 74 and 76) cannot provide an inventive concept. Further, the additional limitation in the claims noted above are directed towards insignificant solution activity. The claims are not patent eligible.
Dependent claims 2-7, 9-14 and 16-20 are also directed toward an abstract idea and do not include additional elements that are sufficient to amount to significantly more than the judicial exception because the additional elements when considered both individually and as an ordered combination do not amount to significantly more than the abstract idea. Therefore, claims 1-20 do not contain patent eligible subject matter that has been identified by the courts.
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.
Claims 1-2, 4, 7-8, 10-12, 14-16 and 19-20 are rejected under 35 U.S.C. 103 as being unpatentable over Churgin et al. Pat No. US 12475884 B2 (Churgin) in view of Cetoli Pat No US 12222992 B1 (Cetoli).
Regarding Claim 1, Churgin discloses a method, comprising:
receiving, by a device (Churgin, col 19, ln 39-40, CSR computer system 700…), transcripts associated with calls or chats (Churgin, col 19, ln 50-52, collection of call record data objects organized by a unique call record identifier and indexing call audio data 784 and related metadata. In some configurations, databases 780 may include call transcript data 786, annotated transcript data 788);
filtering, by the device, customer utterances from the transcripts (Churgin, col 9, ln 67, filtering the utterances in the text transcript);
predicting, by the device, intents for the customer utterances (Churgin, col 1, ln 33-35, a trained intent model is used to process utterances from a call transcript to determine an intent label for each utterance), wherein the intents are associated with corresponding probabilities (Churgin, col 9, ln 9-19, The intent label with the highest confidence score and/or meeting an intent confidence threshold may be assigned that intent label and a corresponding intent label tag may be added to annotated transcript data object 152 in association with the text of that utterance and/or added to other corresponding metadata files or databases);
filtering, by the device, the intents to remove intents with probabilities below a threshold and to generate a filtered set of intents (Churgin, col 9, ln 66 - col 10, ln 15, call summarization engine 154 may include logic for filtering the utterances in the text transcript based on the content tags, such as intent tags, topic tags, and/or sentiment tags. In some configurations, a set of heuristics may be iteratively determined for selecting utterances that constitute the most important moments of the call. For example, a first set of intent filters may be applied to remove utterances unlikely to contribute to the substance of the call, such as greetings, automated prompts, affirmations, abstentions, etc. In some configurations, a key set of intent labels may be identified, such as member questions, CSR answers, and next steps, and all other intent labels may be filtered out. The resulting filtered set of utterances may be stored as an extractive summary file or data object comprised of the subset of utterances and their corresponding speaker, intent, and/or other tags (e.g., topic, sentiment, etc.));
generating, by the device, features from the customer utterances and the filtered set of intents using a natural language processing technique (Churgin, col 8, ln 22-38, natural language processing engine 150 may use a combination of general natural language processing models and natural language processing models trained on domain-specific data to determine the intent of each utterance and one or more topics for each utterance and/or the call as a whole. Natural language processing engine 150 may output an annotated transcript data object 152 that comprises both the utterance text and speaker tags from transcript data object 148 and additional tags for utterance intent, call topics, and/or other metadata describing the content of the text… Annotated transcript data object 152 may then be passed to call summarization engine 150 for extractive filtering of utterances and generation of a natural language summary; Churgin, col 9, ln 22-38, Call summarization engine 154 may include one or more trained natural language processing models configured to process filtered text data from annotated transcript data object 152 to generate a natural language call summary and store it as a natural language summary object 156);
Churgin does not specifically disclose assigning, by the device, respective weights to the filtered set of intents based on a comparison between the features and the customer utterances, ranking, by the device, the filtered set of intents based on the respective weights to identify a primary intent and a secondary intent of one of the calls or the chats, identifying, by the device, a primary intent reason for the primary intent and a secondary intent reason for the secondary intent based on historical data and an association model, and performing, by the device, one or more actions based on the primary intent reason or the secondary intent reason.
However, Cetoli, in the same field of endeavor, discloses:
assigning, by the device, respective weights to the filtered set of intents based on a comparison between the features and the customer utterances (Cetoli, col 4, ln 19 – 24, a distance metric value can be determined between its vector representation and that of the request, and the value is used to assign a ranking. Next, a model of the set of models (same or different model) classifies the output generation request and each chunk with categorical labels (e.g., intent) that indicate attributes of the expected output; [i.e., “distance metric value to be determined by a difference and used to assign ranking” as a weight] );
ranking, by the device, the filtered set of intents based on the respective weights to identify a primary intent and a secondary intent of one of the calls or the chats (Cetoli, col 12, ln 60 – col 14, ln 37, For each particular chunk in the set of chunks, the intent-based data generation platform 104 can determine a distance metric value between the vector representation of the received output generation request and the vector representation of the particular chunk to measure the similarity/dissimilarity of the vectors. Distance metrics can include Euclidean distance… The intent-based data generation platform 104 can use the determined distance metric value between the vector representation of the received output generation request and the vector representation of the particular chunk to assign the particular chunk a ranking within the first set of rankings… the intent-based data generation platform 104 can classify one or more chunks of the set of chunks with multiple categorical labels (e.g., using all labels above a certain confidence threshold). The intent-based data generation platform 104 can generate the ranking of the second set of rankings of the one or more chunks using a weighted sum of the multiple categorical labels; [The intent-based data generation platform 104 assign “a ranking within the first set of rankings” as “primary intent”; “the ranking of the second set of rankings” as “secondary intent”]);
identifying, by the device, a primary intent reason for the primary intent and a secondary intent reason for the secondary intent based on historical data and an association model (Cetoli, col 4, ln 50-52, intent-based rankings identify the context and purpose behind the query; Cetoli, col 7, ln 9-16, Intent ranking engine 114 ranks the classified information 112 based on the inferred intent of the user query 102. Intent ranking engine 114 ensures that the information presented to the user is not only relevant semantically but also aligned with the user's underlying intent, providing a more personalized information retrieval. Intent, in the context of information retrieval and NLP, can refer to the underlying purpose or goal behind a user's query 102 or input; Cetoli, col 13, ln 6- col 14, ln 39, The intent-based data generation platform 104 can use the determined distance metric value between the vector representation of the received output generation request and the vector representation of the particular chunk to assign the particular chunk a ranking within the first set of rankings… The second AI model can use a pre-trained transformer-based architecture, such as BERT or GPT, that has been tuned on a labeled dataset containing examples of text and their corresponding categorical labels to identify the associations between specific phrases, keywords, and the overall context of the text with the predefined labels… The model can further refine its identification of potential categories by using additional contextual information, such as metadata indicating the document's source or historical data on similar queries… The intent-based data generation platform 104 can generate the ranking of the second set of rankings of the one or more chunks using a weighted sum of the multiple categorical labels. Chunks with higher weighted sums are ranked higher, and can indicate greater relevance to the request; [i.e., goal/purpose for first set of intents as “primary intent reason” and goal/purpose for secondary set of intents as “a secondary intent reason”]); and
performing, by the device, one or more actions based on the primary intent reason or the secondary intent reason (Cetoli, col 7, ln 14-50, Intent, in the context of information retrieval and NLP, can refer to the underlying purpose or goal behind a user's query 102 or input and/or the retrieved documents, and encapsulate the user's desired outcome and/or the specific information the user is seeking to obtain. In some implementations, intent can be separated into one or more categories/classifications.…Transactional intent encompasses queries where the user intends to perform a specific action (e.g., making a purchase, booking a service, completing a task). For example, “Open a bank account,” can indicate that the user is ready to engage in a transaction or initiate a service. Investigation intent can include queries where the user is researching products or services (e.g., with the intention of making a purchase decision). Investigation intent can be characterized by the user's need to gather information to inform their decision-making process. For example, “Best financial software for small businesses” can be categorized under investigation intent where the user is evaluating options before making a commitment. One of skill in the art would understand that the disclosed categories of intent is non-limiting, and would understand that the disclosed techniques performed by the intent-based data generation platform 104 described herein can apply to other types of intent).
Therefore, it would have been obvious for one having ordinary skill in the art before the effective filing date of the claimed invention to incorporate the method of Cetoli in the method of Churgin because this would enable intent-based rankings to identify the context and purpose behind the query using LLMs unlike keyword matching, which simply looks for the presence of specific words, and this introduces a natural language processing task of information retrieval (IR) that can be used to generate responses of LLMs and has the task of identifying and retrieving information system resources that are relevant as well as pertaining to an information need (Cetoli, col 3, ln 16-18 and col 1, ln 24-28).
Regarding Claim 2, Churgin in view of Cetoli disclose the method of claim 1, wherein generating the features from the customer utterances and the filtered set of intents (Churgin, col 3, ln 17-28, generate a feature vector for each utterance of the plurality of utterances; process each feature vector through a trained intent model to assign an intent label to each utterance; filter, based on a filter set of intent labels, the plurality of utterances to generate an extractive summary data object, where the extractive summary data object may include a subset of the plurality of utterances and corresponding intent labels and speaker identifiers for the subset of the plurality of utterances; and generate, based on the extractive summary data object and a trained summarization model, a natural language summary data object for the customer service call) comprises:
utilizing a model to generate the features from the customer utterances and the filtered set of intents (Churgin, col 1, ln 44-48, generating a feature vector for each utterance of the plurality of utterances; processing each feature vector through a trained intent model to assign an intent label to each utterance; filtering, based on a filter set of intent labels).
Regarding Claim 4, Churgin in view of Cetoli disclose the method of claim 1, wherein predicting the intents for the customer utterances comprises:
utilizing a small language model to predict the intents for the customer utterances (Churgin, col 1, ln 33-47, a trained intent model is used to process utterances from a call transcript to determine an intent label for each utterance… processing each feature vector through a trained intent model to assign an intent label to each utterance…; Churgin, col 21, ln 5-24, a trained intent model 736 may process the feature vectors for each utterance to classify each utterance according to the speaker's intent. For example, trained intent model 736 may include a multi-class classifier trained for a master set of utterance intent labels 738 and may use the classifier output to assign a specific intent label to each utterance… Natural language processing engine 730 may output an annotated transcript data object comprising content tags (in addition to the utterance boundaries and speaker tags in the original transcript).
Regarding Claim 7, Churgin in view of Cetoli disclose the method of claim 1, wherein filtering the customer utterances from the transcripts comprises:
removing agent utterances from the transcripts (Churgin, col 10, ln 5-7, a first set of intent filters may be applied to remove utterances unlikely to contribute to the substance of the call, such as greetings, automated prompts, affirmations, abstentions, etc.).
Regarding Claim 8, A device, comprising:
one or more processors configured to (Churgin, col 3, ln 14, cause the one or more processors to):receive transcripts associated with calls or chats (Churgin, col 19, ln 39-40, CSR computer system 700…), transcripts associated with calls or chats (Churgin, col 19, ln 50-52, collection of call record data objects organized by a unique call record identifier and indexing call audio data 784 and related metadata. In some configurations, databases 780 may include call transcript data 786, annotated transcript data 788);
filter customer utterances from the transcripts by removing agent utterances from the transcripts (Churgin, col 8, ln 3-37, a custom classifier may be trained to assign speaker labels (CSR, customer/member, etc.) … each speaker is a member, agent, care manager, interactive voice response prompt, etc. … Based on the output of the classifier model, each speaker may be tagged with an appropriate label in transcript data object 148. Natural language processing engine 150 may include one or more trained natural language processing models configured to process text data from transcript data object 148 to identify and label the content of that text… Annotated transcript data object 152 may then be passed to call summarization engine 150 for extractive filtering of utterances; Churgin, col 9, ln 67- col 10, ln 15, Call summarization engine 154 may include one or more trained natural language processing models configured to process filtered text data from annotated transcript data object 152 … call summarization engine 154 may include logic for filtering the utterances in the text transcript… a first set …filters may be applied to remove utterances unlikely to contribute to the substance of the call, such as greetings, automated prompts, affirmations, abstentions, etc. In some configurations, a key set of intent labels may be identified, such as member questions, CSR answers, and next steps, and all other intent labels may be filtered out. The resulting filtered set of utterances may be stored as an extractive summary file or data object comprised of the subset of utterances and their corresponding speaker, intent, and/or other tags (e.g., topic, sentiment, etc.); [i.e., “each speaker may be tagged with an appropriate label in transcript data object… extractive summary file or data object comprised of the subset of utterances and their corresponding speaker”, i.e., the specific user/customer/speaker summary transcript obtained by removing the agent and any other speaker utterances]);
predict intents for the customer utterances, wherein the intents are associated with corresponding probabilities (Churgin, col 1, ln 33-35, a trained intent model is used to process utterances from a call transcript to determine an intent label for each utterance), wherein the intents are associated with corresponding probabilities (Churgin, col 9, ln 9-19, The intent label with the highest confidence score and/or meeting an intent confidence threshold may be assigned that intent label and a corresponding intent label tag may be added to annotated transcript data object 152 in association with the text of that utterance and/or added to other corresponding metadata files or databases);
filter the intents to remove intents with probabilities below a threshold and to generate a filtered set of intents (Churgin, col 9, ln 66 - col 10, ln 15, call summarization engine 154 may include logic for filtering the utterances in the text transcript based on the content tags, such as intent tags, topic tags, and/or sentiment tags. In some configurations, a set of heuristics may be iteratively determined for selecting utterances that constitute the most important moments of the call. For example, a first set of intent filters may be applied to remove utterances unlikely to contribute to the substance of the call, such as greetings, automated prompts, affirmations, abstentions, etc. In some configurations, a key set of intent labels may be identified, such as member questions, CSR answers, and next steps, and all other intent labels may be filtered out. The resulting filtered set of utterances may be stored as an extractive summary file or data object comprised of the subset of utterances and their corresponding speaker, intent, and/or other tags (e.g., topic, sentiment, etc.));
generate features from the customer utterances and the filtered set of intents using a natural language processing technique (Churgin, col 8, ln 22-38, natural language processing engine 150 may use a combination of general natural language processing models and natural language processing models trained on domain-specific data to determine the intent of each utterance and one or more topics for each utterance and/or the call as a whole. Natural language processing engine 150 may output an annotated transcript data object 152 that comprises both the utterance text and speaker tags from transcript data object 148 and additional tags for utterance intent, call topics, and/or other metadata describing the content of the text… Annotated transcript data object 152 may then be passed to call summarization engine 150 for extractive filtering of utterances and generation of a natural language summary; Churgin, col 9, ln 22-38, Call summarization engine 154 may include one or more trained natural language processing models configured to process filtered text data from annotated transcript data object 152 to generate a natural language call summary and store it as a natural language summary object 156);
Churgin does not specifically disclose assign respective weights to the filtered set of intents based on a comparison between the features and the customer utterances, rank the filtered set of intents based on the respective weights to identify a primary intent and a secondary intent of one of the calls or the chats, identify a primary intent reason for the primary intent and a secondary intent reason for the secondary intent based on historical data and an association model, and perform one or more actions based on the primary intent reason or the secondary intent reason.
However, Cetoli, in the same field of endeavor, discloses:
assign respective weights to the filtered set of intents based on a comparison between the features and the customer utterances (Cetoli, col 4, ln 19 – 24, a distance metric value can be determined between its vector representation and that of the request, and the value is used to assign a ranking. Next, a model of the set of models (same or different model) classifies the output generation request and each chunk with categorical labels (e.g., intent) that indicate attributes of the expected output; [i.e., “distance metric value to be determined by a difference and used to assign ranking” as a weight]);
rank the filtered set of intents based on the respective weights to identify a primary intent and a secondary intent of one of the calls or the chats (Cetoli col 12, ln 60 – col 14, ln 37, For each particular chunk in the set of chunks, the intent-based data generation platform 104 can determine a distance metric value between the vector representation of the received output generation request and the vector representation of the particular chunk to measure the similarity/dissimilarity of the vectors. Distance metrics can include Euclidean distance… The intent-based data generation platform 104 can use the determined distance metric value between the vector representation of the received output generation request and the vector representation of the particular chunk to assign the particular chunk a ranking within the first set of rankings … the intent-based data generation platform 104 can classify one or more chunks of the set of chunks with multiple categorical labels (e.g., using all labels above a certain confidence threshold). The intent-based data generation platform 104 can generate the ranking of the second set of rankings of the one or more chunks using a weighted sum of the multiple categorical labels; [The intent-based data generation platform 104 assign “a ranking within the first set of rankings” as “primary intent”; “the ranking of the second set of rankings” as “secondary intent”]);
identify a primary intent reason for the primary intent and a secondary intent reason for the secondary intent based on historical data and an association model (Cetoli, col 4, ln 50-52, intent-based rankings identify the context and purpose behind the query; Cetoli, col 7, ln 9-16, Intent ranking engine 114 ranks the classified information 112 based on the inferred intent of the user query 102. Intent ranking engine 114 ensures that the information presented to the user is not only relevant semantically but also aligned with the user's underlying intent, providing a more personalized information retrieval. Intent, in the context of information retrieval and NLP, can refer to the underlying purpose or goal behind a user's query 102 or input; Cetoli, col 13, ln 6- col 14, ln 39, The intent-based data generation platform 104 can use the determined distance metric value between the vector representation of the received output generation request and the vector representation of the particular chunk to assign the particular chunk a ranking within the first set of rankings… The second AI model can use a pre-trained transformer-based architecture, such as BERT or GPT, that has been tuned on a labeled dataset containing examples of text and their corresponding categorical labels to identify the associations between specific phrases, keywords, and the overall context of the text with the predefined labels… The model can further refine its identification of potential categories by using additional contextual information, such as metadata indicating the document's source or historical data on similar queries… The intent-based data generation platform 104 can generate the ranking of the second set of rankings of the one or more chunks using a weighted sum of the multiple categorical labels. Chunks with higher weighted sums are ranked higher, and can indicate greater relevance to the request; [i.e., goal/purpose for first set of intents as “primary intent reason” and goal/purpose for secondary set of intents as “a secondary intent reason”]); and
perform one or more actions based on the primary intent reason or the secondary intent reason (Cetoli, col 7, ln 14-50, Intent, in the context of information retrieval and NLP, can refer to the underlying purpose or goal behind a user's query 102 or input and/or the retrieved documents, and encapsulate the user's desired outcome and/or the specific information the user is seeking to obtain. In some implementations, intent can be separated into one or more categories/classifications.…Transactional intent encompasses queries where the user intends to perform a specific action (e.g., making a purchase, booking a service, completing a task). For example, “Open a bank account,” can indicate that the user is ready to engage in a transaction or initiate a service. Investigation intent can include queries where the user is researching products or services (e.g., with the intention of making a purchase decision). Investigation intent can be characterized by the user's need to gather information to inform their decision-making process. For example, “Best financial software for small businesses” can be categorized under investigation intent where the user is evaluating options before making a commitment. One of skill in the art would understand that the disclosed categories of intent is non-limiting, and would understand that the disclosed techniques performed by the intent-based data generation platform 104 described herein can apply to other types of intent).
Therefore, it would have been obvious for one having ordinary skill in the art before the effective filing date of the claimed invention to incorporate the method of Cetoli in the method of Churgin because this would enable intent-based rankings to identify the context and purpose behind the query using LLMs unlike keyword matching, which simply looks for the presence of specific words, and this introduces a natural language processing task of information retrieval (IR) that can be used to generate responses of LLMs and has the task of identifying and retrieving information system resources that are relevant as well as pertaining to an information need (Cetoli, col 3, ln 16-18 and col 1, ln 24-28).
Regarding Claim 10, Churgin in view of Cetoli disclose the device of claim 8, wherein the one or more processors are further configured to:
Cetoli further discloses:
aggregate the customer utterances into dynamic windows prior to predicting the intents for the customer utterances (Cetoli, , For audio, the intent-based data generation platform 104 can use techniques such as Mel-frequency cepstral coefficients (MFCCs) or spectrograms to extract features (e.g., pitch, formants, zero-crossing rates) from audio signals. The features can be indexed in a database, and for a given query, the intent-based data generation platform 104 can use the extracted features from the query audio to perform a similarity search to map audio files. Multi-modal output generation requests, which can include a combination of text, images, videos, and/or audio, can be separated by the mode, and the intent-based data generation platform 104 can extract and index features from each modality separately. For a given query, the intent-based data generation platform 104 can extract features from each modality, combine them into representative vectors (e.g., weighing each modality equally or differently), and perform a similarity search to retrieve relevant multi-modal documents ;).
Regarding Claim 11, Churgin in view of Cetoli disclose the device of claim 8, wherein the one or more processors.
utilize a similarity analysis to compare the features with the customer utterances to determine similarity scores (Churgin, col 8, ln 11, ln 24 – 40, each utterance may be isolated and embedded for processing through one or more natural language processing models. For example, a natural language processing engine may use each utterance defined in the tagging of the text transcript and generate corresponding numerical value vectors in a lower-dimensional space that represents similar words as similar values… specific configurations of feature vectors may be generated for one or more natural language processing models. For example, the natural language processing engine may use a character-based BERT deep learning model to convert each utterance into a dense feature vector to support intent classification); and
assign the respective weights to the filtered set of intents based on the similarity scores and predefined weighting factors (Churgin, col 8, ln 58 - col 9, ln 33, an utterance intent classifier model may be trained using weak supervision to provide an unsupervised custom classifier of intents for the specific domain. In some configurations, a set of heuristics for weak supervision may be iteratively defined for customer service calls and, more specifically, a specialized domain, such as medical services, pharmacy, or insurance, to define a set of intent labels to apply to each utterance or segment of the call, such as “greeting”, “member question”, “representative answer”, “follow-up question”, etc…The corresponding topic label assignment logic may also use a topic confidence threshold to identify more than one topic and/or weighted topic values for the topic labels; Churgin, col 10, ln 4-11, …a first set of intent filters may be applied to remove utterances unlikely to contribute to the substance of the call, such as greetings, automated prompts, affirmations, abstentions, etc. In some configurations, a key set of intent labels may be identified, such as member questions, CSR answers, and next steps, and all other intent labels may be filtered out).
Cetoli further discloses:
to assign the respective weights to the filtered set of intents based on the comparison between the features and the customer utterances (Cetoli, col 4, ln 19 – 24, a distance metric value can be determined between its vector representation and that of the request, and the value is used to assign a ranking. Next, a model of the set of models (same or different model) classifies the output generation request and each chunk with categorical labels (e.g., intent) that indicate attributes of the expected output; [i.e., “distance metric value to be determined by a difference and used to assign ranking” as a weight] ).
Regarding Claim 12, Churgin in view of Cetoli disclose the device of claim 11, wherein the one or more processors, to utilize the similarity analysis to compare the features with the customer utterances to determine the similarity scores, are configured to:
Cetoli further discloses:
utilize cosine similarity to determine the similarity scores based on degrees of similarity between the features and the customer utterances (Cetoli, col 11, ln 28 – col 13, ln 14, the intent-based data generation platform 104 can perform feature extraction using pre-trained CNNs. The extracted feature vectors can be indexed in a database (e.g., a vector database) for subsequent similarity searches. When a query is received, the intent-based data generation platform 104 can extract the feature vector of the query image using the same CNN model and transformations, and performs a similarity search using metrics like cosine similarity or Euclidean distance to find similar images in the database… In operation 308, the intent-based data generation platform 104 can generate, by the first AI model, a first set of rankings of the set of chunks. For example, the intent-based data generation platform 104 can generate a set of vector representations of (i) the received output generation request and/or (ii) each chunk of the set of chunks of the retrieved set of documents. The intent-based data generation platform 104 can use a pre-trained transformer-based model, such as Bidirectional Encoder Representations from Transformers (BERT) or Generative Pre-trained Transformer (GPT). The chunks can be fed into the AI model to produce vector representations. For each particular chunk in the set of chunks, the intent-based data generation platform 104 can determine a distance metric value between the vector representation of the received output generation request and the vector representation of the particular chunk to measure the similarity/dissimilarity of the vectors… The intent-based data generation platform 104 can use the determined distance metric value between the vector representation of the received output generation request and the vector representation of the particular chunk to assign the particular chunk a ranking within the first set of rankings. Chunks with vector representations that are closer to the vector representation of the output generation request or have a higher cosine similarity (indicating higher relevance) receive higher rankings; [i.e., “audio” as “customer utterances”; “assign the particular chunk a ranking within the first set of rankings. Chunks with vector representations that are closer to the vector representation of the output generation request or have a higher cosine similarity (indicating higher relevance) receive higher rankings” as “utilize cosine similarity to determine the similarity scores based on degrees of similarity”]).
Regarding Claim 14, Churgin in view of Cetoli disclose the device of claim 8, wherein the one or more processors, to perform the one or more actions based on the primary intent reason or the secondary intent reason, are configured to one or more of:
provide the primary intent reason or the secondary intent reason for display;
Cetoli further discloses:
identify an originating factor of the one of the calls or the chats based on the primary intent reason or the secondary intent reason (Cetoli, col 3, ln 16 - col 4, ln 33, intent-based rankings identify the context and purpose behind the query… The model generates a first set of rankings for the chunks by creating vector representations of both the output generation request and each chunk. For each chunk, a distance metric value can be determined between its vector representation and that of the request, and the value is used to assign a ranking… The model generates a second set of rankings for the chunks based on the categorical labels, ranking chunks with matching labels higher than those without matching labels; [i.e., “intent-based rankings identify the context and purpose … first set of rankings” as “first/primary reason”; “intent-based rankings identify the context and purpose … second set of rankings” as “secondary reason”;]);
address a customer issue based on the primary intent reason or the secondary intent reason;
generate a business metric based on the primary intent reason or the secondary intent reason; or
retrain the association model based on the primary intent reason or the secondary intent reason.
Regarding Claim 15, A non-transitory computer-readable medium storing a set of instructions, the set of instructions (Churgin, col 26, ln 7-10, computer readable medium can be any non-transitory storage apparatus that can contain, store, communicate, propagate, or transport the program for use by or in connection with the instruction execution system) comprising:
one or more instructions that, when executed by one or more processors of a device, cause the device to (Churgin, col 26, ln 13-15, executing program code may include at least one processor coupled directly or indirectly to memory elements through a system bus):
receive transcripts associated with calls or chats (Churgin, col 19, ln 50-52, collection of call record data objects organized by a unique call record identifier and indexing call audio data 784 and related metadata. In some configurations, databases 780 may include call transcript data 786, annotated transcript data 788);
filter customer utterances from the transcripts (Churgin, col 9, ln 67, filtering the utterances in the text transcript);
utilize a small language model to predict intents for the customer utterances (Churgin, col 1, ln 33-35, a trained intent model is used to process utterances from a call transcript to determine an intent label for each utterance), wherein the intents are associated with corresponding probabilities (Churgin, col 9, ln 9-19, The intent label with the highest confidence score and/or meeting an intent confidence threshold may be assigned that intent label and a corresponding intent label tag may be added to annotated transcript data object 152 in association with the text of that utterance and/or added to other corresponding metadata files or databases);
filter the intents to remove intents with probabilities below a threshold and to generate a filtered set of intents (Churgin, col 9, ln 66 - col 10, ln 15, call summarization engine 154 may include logic for filtering the utterances in the text transcript based on the content tags, such as intent tags, topic tags, and/or sentiment tags. In some configurations, a set of heuristics may be iteratively determined for selecting utterances that constitute the most important moments of the call. For example, a first set of intent filters may be applied to remove utterances unlikely to contribute to the substance of the call, such as greetings, automated prompts, affirmations, abstentions, etc. In some configurations, a key set of intent labels may be identified, such as member questions, CSR answers, and next steps, and all other intent labels may be filtered out. The resulting filtered set of utterances may be stored as an extractive summary file or data object comprised of the subset of utterances and their corresponding speaker, intent, and/or other tags (e.g., topic, sentiment, etc.));
generate features from the customer utterances and the filtered set of intents using a natural language processing technique (Churgin, col 8, ln 22-38, natural language processing engine 150 may use a combination of general natural language processing models and natural language processing models trained on domain-specific data to determine the intent of each utterance and one or more topics for each utterance and/or the call as a whole. Natural language processing engine 150 may output an annotated transcript data object 152 that comprises both the utterance text and speaker tags from transcript data object 148 and additional tags for utterance intent, call topics, and/or other metadata describing the content of the text… Annotated transcript data object 152 may then be passed to call summarization engine 150 for extractive filtering of utterances and generation of a natural language summary; Churgin, col 9, ln 22-38, Call summarization engine 154 may include one or more trained natural language processing models configured to process filtered text data from annotated transcript data object 152 to generate a natural language call summary and store it as a natural language summary object 156); Churgin does not specifically disclose assign respective weights to the filtered set of intents based on a comparison between the features and the customer utterances, rank the filtered set of intents based on the respective weights to identify a primary intent and a secondary intent of one of the calls or the chats, identify a primary intent reason for the primary intent and a secondary intent reason for the secondary intent based on historical data and an association model, and perform one or more actions based on the primary intent reason or the secondary intent reason.
However, Cetoli, in the same field of endeavor, discloses:
assign respective weights to the filtered set of intents based on a comparison between the features and the customer utterances (Cetoli, col 4, ln 19 – 24, a distance metric value can be determined between its vector representation and that of the request, and the value is used to assign a ranking. Next, a model of the set of models (same or different model) classifies the output generation request and each chunk with categorical labels (e.g., intent) that indicate attributes of the expected output; [i.e., “distance metric value to be determined by a difference and used to assign ranking” as a weight] );
rank the filtered set of intents based on the respective weights to identify a primary intent and a secondary intent of one of the calls or the chats (Cetoli col 12, ln 60 – col 14, ln 37, For each particular chunk in the set of chunks, the intent-based data generation platform 104 can determine a distance metric value between the vector representation of the received output generation request and the vector representation of the particular chunk to measure the similarity/dissimilarity of the vectors. Distance metrics can include Euclidean distance… The intent-based data generation platform 104 can use the determined distance metric value between the vector representation of the received output generation request and the vector representation of the particular chunk to assign the particular chunk a ranking within the first set of rankings … the intent-based data generation platform 104 can classify one or more chunks of the set of chunks with multiple categorical labels (e.g., using all labels above a certain confidence threshold). The intent-based data generation platform 104 can generate the ranking of the second set of rankings of the one or more chunks using a weighted sum of the multiple categorical labels; [The intent-based data generation platform 104 assign “a ranking within the first set of rankings” as “primary intent”; “the ranking of the second set of rankings” as “secondary intent”]);
identify a primary intent reason for the primary intent and a secondary intent reason for the secondary intent based on historical data and an association model (Cetoli, col 4, ln 50-52, intent-based rankings identify the context and purpose behind the query; Cetoli, col 7, ln 9-16, Intent ranking engine 114 ranks the classified information 112 based on the inferred intent of the user query 102. Intent ranking engine 114 ensures that the information presented to the user is not only relevant semantically but also aligned with the user's underlying intent, providing a more personalized information retrieval. Intent, in the context of information retrieval and NLP, can refer to the underlying purpose or goal behind a user's query 102 or input; Cetoli, col 13, ln 6- col 14, ln 39, The intent-based data generation platform 104 can use the determined distance metric value between the vector representation of the received output generation request and the vector representation of the particular chunk to assign the particular chunk a ranking within the first set of rankings… The second AI model can use a pre-trained transformer-based architecture, such as BERT or GPT, that has been tuned on a labeled dataset containing examples of text and their corresponding categorical labels to identify the associations between specific phrases, keywords, and the overall context of the text with the predefined labels… The model can further refine its identification of potential categories by using additional contextual information, such as metadata indicating the document's source or historical data on similar queries… The intent-based data generation platform 104 can generate the ranking of the second set of rankings of the one or more chunks using a weighted sum of the multiple categorical labels. Chunks with higher weighted sums are ranked higher, and can indicate greater relevance to the request; [i.e., goal/purpose for first set of intents as “primary intent reason” and goal/purpose for secondary set of intents as “a secondary intent reason”]); and
perform one or more actions based on the primary intent reason or the secondary intent reason (Cetoli, col 7, ln 14-50, Intent, in the context of information retrieval and NLP, can refer to the underlying purpose or goal behind a user's query 102 or input and/or the retrieved documents, and encapsulate the user's desired outcome and/or the specific information the user is seeking to obtain. In some implementations, intent can be separated into one or more categories/classifications.…Transactional intent encompasses queries where the user intends to perform a specific action (e.g., making a purchase, booking a service, completing a task). For example, “Open a bank account,” can indicate that the user is ready to engage in a transaction or initiate a service. Investigation intent can include queries where the user is researching products or services (e.g., with the intention of making a purchase decision). Investigation intent can be characterized by the user's need to gather information to inform their decision-making process. For example, “Best financial software for small businesses” can be categorized under investigation intent where the user is evaluating options before making a commitment. One of skill in the art would understand that the disclosed categories of intent is non-limiting, and would understand that the disclosed techniques performed by the intent-based data generation platform 104 described herein can apply to other types of intent).
Therefore, it would have been obvious for one having ordinary skill in the art before the effective filing date of the claimed invention to incorporate the method of Cetoli in the method of Churgin because this would enable intent-based rankings to identify the context and purpose behind the query using LLMs unlike keyword matching, which simply looks for the presence of specific words, and this introduces a natural language processing task of information retrieval (IR) that can be used to generate responses of LLMs and has the task of identifying and retrieving information system resources that are relevant as well as pertaining to an information need (Cetoli, col 3, ln 16-18 and col 1, ln 24-28).
Regarding Claim 16, Churgin in view of Cetoli disclose the non-transitory computer-readable medium of claim 15, wherein the one or more
instructions, that cause the device to generate the features from the customer utterances and the filtered set of intents, cause the device (Churgin, col 3, ln 54-col 4, ln 2, determine a plurality of training utterances and corresponding feature vectors for the plurality of call transcript data objects…generate a master set of intent labels for the trained intent model, where the filter set of intent labels is a selected subset of the master set of intent labels; The instructions may further include instructions to: generate, from the natural language summary data object, an embedding and at least one corresponding feature vector for the natural language summary data object) to:
utilize a model to generate the features from the customer utterances and the filtered set of intents (Churgin, col 2, ln 18 – 28, processing the plurality of training utterances and corresponding feature vectors through the trained intent model to determine weighting coefficients for the trained intent model and at least one model acceptance metric; and generating a master set of intent labels for the trained intent model, where the filter set of intent labels is a selected subset of the master set of intent labels. The computer-implemented method may include: generating, from the natural language summary data object, an embedding and at least one corresponding feature vector for the natural language summary data object).
Regarding Claim 19, Churgin in view of Cetoli disclose the non-transitory computer-readable medium of claim 15, wherein the one or more instructions further cause the device to:
Cetoli further discloses:
aggregate the customer utterances into dynamic windows prior to predicting the intents for the customer utterances (Cetoli, For audio, the intent-based data generation platform 104 can use techniques such as Mel-frequency cepstral coefficients (MFCCs) or spectrograms to extract features (e.g., pitch, formants, zero-crossing rates) from audio signals. The features can be indexed in a database, and for a given query, the intent-based data generation platform 104 can use the extracted features from the query audio to perform a similarity search to map audio files. Multi-modal output generation requests, which can include a combination of text, images, videos, and/or audio, can be separated by the mode, and the intent-based data generation platform 104 can extract and index features from each modality separately. For a given query, the intent-based data generation platform 104 can extract features from each modality, combine them into representative vectors (e.g., weighing each modality equally or differently), and perform a similarity search to retrieve relevant multi-modal documents ).
Regarding Claim 20, Churgin in view of Cetoli disclose the non-transitory computer-readable medium of claim 15.
utilize a similarity analysis to compare the features with the customer utterances to determine similarity scores (Churgin, col 8, ln 11, ln 24 – 40, each utterance may be isolated and embedded for processing through one or more natural language processing models. For example, a natural language processing engine may use each utterance defined in the tagging of the text transcript and generate corresponding numerical value vectors in a lower-dimensional space that represents similar words as similar values… specific configurations of feature vectors may be generated for one or more natural language processing models. For example, the natural language processing engine may use a character-based BERT deep learning model to convert each utterance into a dense feature vector to support intent classification); and
assign the respective weights to the filtered set of intents based on the similarity scores and predefined weighting factors (Churgin, col 8, ln 58 - col 9, ln 33, an utterance intent classifier model may be trained using weak supervision to provide an unsupervised custom classifier of intents for the specific domain. In some configurations, a set of heuristics for weak supervision may be iteratively defined for customer service calls and, more specifically, a specialized domain, such as medical services, pharmacy, or insurance, to define a set of intent labels to apply to each utterance or segment of the call, such as “greeting”, “member question”, “representative answer”, “follow-up question”, etc…The corresponding topic label assignment logic may also use a topic confidence threshold to identify more than one topic and/or weighted topic values for the topic labels; Churgin, col 10, ln 4-11, …a first set of intent filters may be applied to remove utterances unlikely to contribute to the substance of the call, such as greetings, automated prompts, affirmations, abstentions, etc. In some configurations, a key set of intent labels may be identified, such as member questions, CSR answers, and next steps, and all other intent labels may be filtered out).
Cetoli further discloses:
wherein the one or more instructions, that cause the device to assign the respective weights to the filtered set of intents based on the comparison between the features and the customer utterances (Cetoli, col 4, ln 19 – 24, a distance metric value can be determined between its vector representation and that of the request, and the value is used to assign a ranking. Next, a model of the set of models (same or different model) classifies the output generation request and each chunk with categorical labels (e.g., intent) that indicate attributes of the expected output; [i.e., “distance metric value to be determined by a difference and used to assign ranking” as a weight]).
Claims 3 and 17 are rejected under 35 U.S.C. 103 as being unpatentable over Churgin in view of Cetoli, and further in view of Uma Pat Pub No. US 20260161717 A1 (Uma) (Effective Filing Date: 2024-12-06).
Regarding Claim 3, Churgin in view of Cetoli disclose the method of claim 1, wherein assigning the respective weights to the filtered set of intents comprises:
Churgin in view of Cetoli do not specifically disclose assigning greater weights to intents, of the filtered set of intents, associated with a top feature compared to respective weights assigned to intents, of the filtered set of intents, associated with other features.
However, Uma, in the same field of endeavor, discloses assigning greater weights to intents, of the filtered set of intents, associated with a top feature compared to respective weights assigned to intents, of the filtered set of intents, associated with other features (Uma, para 0114, Only parameters with a support value greater than a predefined threshold are used in the filtering process, meaning that the system gives more weight to the parameters that are most likely to help identify the user's intention).
Therefore, it would have been obvious for one having ordinary skill in the art before the effective filing date of the claimed invention to incorporate the method of Uma in the method of Churgin in view of Cetoli because this would enable systems and methods to provide personalized search query disambiguation by identifying user intentions for ambiguous search terms using user profiles, for instance, considering parameters like Profession, Interests, Gender, and Location represent key attributes that the system considers when disambiguating user queries and each of these parameters is weighted based on its relevance to the user's current query (Uma, Figure 10, para 0114 and Abstract).
Regarding Claim 17, Churgin in view of Cetoli disclose the non-transitory computer-readable medium of claim 15, wherein the one or more instructions, that cause the device to assign the respective weights to the filtered set of intents, cause the device to:
Churgin in view of Cetoli do not specifically disclose assign greater weights to intents, of the filtered set of intents, associated with a top feature compared to respective weights assigned to intents, of the filtered set of intents, associated with other features.
However, Uma, in the same field of endeavor, discloses assign greater weights to intents, of the filtered set of intents, associated with a top feature compared to respective weights assigned to intents, of the filtered set of intents, associated with other features (Uma, para 0114, Only parameters with a support value greater than a predefined threshold are used in the filtering process, meaning that the system gives more weight to the parameters that are most likely to help identify the user's intention).
Therefore, it would have been obvious for one having ordinary skill in the art before the effective filing date of the claimed invention to incorporate the method of Uma in the method of Churgin in view of Cetoli because this would enable systems and methods to provide personalized search query disambiguation by identifying user intentions for ambiguous search terms using user profiles, for instance, considering parameters like Profession, Interests, Gender, and Location represent key attributes that the system considers when disambiguating user queries and each of these parameters is weighted based on its relevance to the user's current query (Uma, Figure 10, para 0114 and Abstract).
Claims 5 are rejected under 35 U.S.C. 103 as being unpatentable over Churgin in view of Cetoli, and further in view of Das et al. Pat App No US 20220309250 A1 (Das).
Regarding Claim 5, Churgin in view of Cetoli disclose the method of claim 1, further comprising:
Churgin in view of Cetoli do not specifically disclose utilizing the historical data to train the association model prior to identifying the primary intent reason for the primary intent and the secondary intent reason for the secondary intent.
However, Das, in the same field of endeavor, discloses utilizing the historical data to train the association model prior to identifying the primary intent reason for the primary intent and the secondary intent reason for the secondary intent (Das, para 0034…0039, The domain data preparation 220 and AI lab 230 stages may relate to decision making in connection with whether to escalate a product support case from the chatbot to a live human agent as well as associated learning from escalated cases. According to one embodiment, historical case data preparation and feature selection and labeling logic may be performed as part of the domain data preparation 220 stage, including preprocessing of case data, including concatenating together the “issue subject” and the “issue description” fields, cleaning (e.g., removal of punctuation, numbers, and stop words), tokenization (e.g., using unigram, bigram, and trigrams). The case history text can then be filtered by product line to create a custom corpus that may be fed as input to a word association model (e.g., Word2Vec)…multiple product line specific word association models 346, which may be trained for each product line based on case history text filtered by product line. For example, the multiple product line specific word association models 346 may be generated based on the results of the historical case data preparation in the domain data preparation 220 stage; Das, para 0013, training AI classification models (without undue upfront manual labeling of training data) to allow the chatbot to accurately identify customers' intent, and mapping of the identified intent to one of the supported product issue categories…).
Therefore, it would have been obvious for one having ordinary skill in the art before the effective filing date of the claimed invention to incorporate the method of Das in the method of Churgin in view of Cetoli because this would enable producing a product line specific word association model for each product line and this provides chatbots more power in interpreting natural language, to both understand better, and learn over time (Das, para 0034 and 0001).
Claim 6 is rejected under 35 U.S.C. 103 as being unpatentable over Churgin in view of Cetoli, and further in view of Matthew et al. Pat App No.AU 2022280709 A1 (Matthew).
Regarding Claim 6, Churgin in view of Cetoli disclose the method of claim 1.
Churgin in view of Cetoli do not specifically disclose wherein the association model generates a reason score for each intent of the filtered set of intents based on a frequency of each intent being associated with the primary intent in the historical data.
However, Matthew, in the same field of endeavor, discloses wherein the association model generates a reason score for each intent of the filtered set of intents based on a frequency of each intent being associated with the primary intent in the historical data (Matthew, para 0035-0036, the intent processing system 102 may identify intent terms and phrases associated with intents commonly submitted by customers of the brand over a period of time or associated with historical intents… The intent processing system 102 may, thus, generate the word cloud 104 based on the frequency of intent terms and/or phrases corresponding to intents of a customer base of the brand or to intents of the customer 108 itself… the intent processing system 102 implements a machine learning algorithm or artificial intelligence that utilizes historical conversation data for the customer 108 and other customers associated…a dataset of input messages and corresponding intents may be analyzed using a clustering algorithm to identify correlations between different types of intent terms and/or phrases with different intents or intent types. Example clustering algorithms that may be trained using sample member attributes and representative attributes (e.g., historical data , hypothetical data, etc.) to identify potential pairings may include a k-means clustering algorithms, fuzzy c-means (FCM) algorithms, expectation- maximization (EM) algorithms, hierarchical clustering algorithms…; Matthew, para 0043-0045, determine a sub-score to one or more factors such as these and a weight to assign to each score. By combining (e.g., summing) weighted sub-scores, a score for each agent 116 can be determined. An agent selection can then be made by comparing agents’ scores (e.g., to select a high or highest score). If the intent processing system 102 determines, based on the identified intent, that another service 112 may provide resolution for the intent, the intent processing system 102 may route communications from the customer 108 to an endpoint corresponding to the other service 112. For example, if the intent corresponds to arranging a payment for a phone bill issued by a wireless carrier service, the intent processing system 102 may route the customer 108 to the wireless carrier service to allow the customer 108, via the computing device 110, to make a payment for the existing phone bill. In some instances, the intent processing system 102 can monitor communications between the customer 108 and an agent 116 or other service 112 to determine a customer sentiment with regard to the identification of the customer’s intent based on the customer’s selections of intent terms and/or phrases from the word cloud 104 and to the routing of communications based on the identification of the customer’s intent. For instance, if the customer 108 expresses in a communication that the intent processing system 102 has selected an undesirable action based on the intent terms and/or phrases selected by the customer 108, the intent processing system 102 may utilize this feedback to retrain the clustering algorithm or artificial intelligence used to associate intent terms and/or phrases with particular intents to more accurately associate the customer’s selections with the appropriate intent; [“the intent processing system 102… utilize this feedback to retrain the clustering algorithm or artificial intelligence used to associate intent” as “the association model”]).
Therefore, it would have been obvious for one having ordinary skill in the art before the effective filing date of the claimed invention to incorporate the method of Matthew in the method of Churgin in view of Cetoli because this would provide a framework to deploy to identify intents based on user inputs and cluster selections and it also provides actionable responses to the identified intents, and this framework has the benefit of making use of a set of intent terms associated with an intent cluster, the set of intent clusters in turn correspond to a set of frequently detected intents, and the set of frequently detected intents are identified based on an evaluation of historical conversation data collected over a period of time (Matthew, para 0002 and 0004-0005).
Claims 9 and 18 are rejected under 35 U.S.C. 103 as being unpatentable over Churgin in view of Cetoli, and further in view of Mangalam et al. Pat App No. US 20250217603 A1 (Mangalam).
Regarding Claim 9, Churgin in view of Cetoli disclose the device of claim 8, wherein the one or more processors are further configured to:
receive a new transcript that includes new customer utterances associated with a new call or a new chat (Churgin, col 6, ln 57- col 7, ln 18, Call processing module 140 may be configured to process call data, such as a recorded call, in order to generate, store, and display call summaries. As described above, audio processing module 142 may receive the audio data from a service call from audio processing module 122. The audio data may be received in real-time and/or following completion of the service call… Call processing module 140 may receive or access audio data object 144 and pass it to a trained speech recognition engine 146 for processing… speech recognition engine 146 may determine separate speakers, such as one (or more) CSRs and one (or more) customers, and convert the audio data corresponding to each speaker into a text transcript data object 148. For example, text transcript data object 148 may include a plurality of transcribed text utterances ascribed to each speaker and organized in the chronological order in which they occurred during the call; [“convert the audio data corresponding to each speaker into a text transcript data object 148…plurality of transcribed text utterances ascribed to each speaker” as “receive a new transcript” for each speaker]); and
Churgin in view of Cetoli do not specifically disclose process the new transcript, with the association model, to identify a new primary intent and a new primary intent reason for the new primary intent.
However, Mangalam, in the same field of endeavor, discloses process the new transcript, with the association model, to identify a new primary intent and a new primary intent reason for the new primary intent (Mangalam, para 0006, the LLM may be prompted to identify the concepts: intent, reason, and action from the call transcript; Mangalam, para 0012, the LLM may identify the intent, reason, and action from an input transcript. The neural networks may each execute to classify its respective concept. For example, the intent classifier may classify the transcript according to an intent, an intent category, an intent sub-category. The reason and action classifiers may similarity classify the transcript. The result is an that the input transcript may be classified in a way that intent-category-sub-category, reason-category-sub-category; Mangalam, para 0068-0070, At 306, the intent categories, actions, and call reasons may be mapped together. For example, the analysis subsystem 138 may identify a top N percentile of intents-to actions-to reasons mappings to discover insights on these data… generating a text embedding of the concept based on the words or phrases that specify the concept, wherein an association is stored between each text embedding and the content from which the concept was identified).
Therefore, it would have been obvious for one having ordinary skill in the art before the effective filing date of the claimed invention to incorporate the method of Matthew in the method of Churgin in view of Cetoli because this would help driving efficiency processes and mitigating technology or other process shortcomings and these insights may include potential issues that are to be mitigated or improved (Mangalam, para 0068).
Regarding Claim 18, Churgin in view of Cetoli disclose the non-transitory computer-readable medium of claim 15, wherein the one or more instructions further cause the device to:
receive a new transcript that includes new customer utterances associated with a new call or a new chat (Churgin, col 6, ln 57- col 7, ln 18, Call processing module 140 may be configured to process call data, such as a recorded call, in order to generate, store, and display call summaries. As described above, audio processing module 142 may receive the audio data from a service call from audio processing module 122. The audio data may be received in real-time and/or following completion of the service call… Call processing module 140 may receive or access audio data object 144 and pass it to a trained speech recognition engine 146 for processing… speech recognition engine 146 may determine separate speakers, such as one (or more) CSRs and one (or more) customers, and convert the audio data corresponding to each speaker into a text transcript data object 148. For example, text transcript data object 148 may include a plurality of transcribed text utterances ascribed to each speaker and organized in the chronological order in which they occurred during the call; [“convert the audio data corresponding to each speaker into a text transcript data object 148…plurality of transcribed text utterances ascribed to each speaker” as “receive a new transcript” for each speaker]); and
Churgin in view of Cetoli do not specifically disclose process the new transcript, with the association model, to identify a new primary intent and a new primary intent reason for the new primary intent.
However, Mangalam, in the same field of endeavor, discloses process the new transcript, with the association model, to identify a new primary intent and a new primary intent reason for the new primary intent (Mangalam, para 0006, the LLM may be prompted to identify the concepts: intent, reason, and action from the call transcript; Mangalam, para 0012, the LLM may identify the intent, reason, and action from an input transcript. The neural networks may each execute to classify its respective concept. For example, the intent classifier may classify the transcript according to an intent, an intent category, an intent sub-category. The reason and action classifiers may similarity classify the transcript. The result is an that the input transcript may be classified in a way that intent-category-sub-category, reason-category-sub-category; Mangalam, para 0068-0070, At 306, the intent categories, actions, and call reasons may be mapped together. For example, the analysis subsystem 138 may identify a top N percentile of intents-to actions-to reasons mappings to discover insights on these data… generating a text embedding of the concept based on the words or phrases that specify the concept, wherein an association is stored between each text embedding and the content from which the concept was identified).
Therefore, it would have been obvious for one having ordinary skill in the art before the effective filing date of the claimed invention to incorporate the method of Matthew in the method of Churgin in view of Cetoli because this would help driving efficiency processes and mitigating technology or other process shortcomings and these insights may include potential issues that are to be mitigated or improved (Mangalam, para 0068).
Allowable Subject Matter
11. Claim 13 is objected to as being dependent upon rejected base claims, but would be allowable if rewritten in independent form including all of the limitations of the base claim and any intervening claims, and also if all these claims overcome the 101 rejections. The reasons for allowance are that the prior art of record do not specifically teach the limitations as recited in the mentioned claims.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to MULUGETA T. DUGDA whose telephone number is (703)756-1106. The examiner can normally be reached Mon - Fri, 4:30am - 7:00pm.
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, Paras D. Shah can be reached at 571-270-1650. 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.
/MULUGETA TUJI DUGDA/Examiner, Art Unit 2653
/Paras D Shah/Supervisory Patent Examiner, Art Unit 2653
07/25/2026
Prosecution Insights