DETAILED ACTION
This communication is in response to the Application filed on 3/24/2023. Claims 1-20 are pending and have been examined.
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 13, 2013, is being examined under the first inventor to file provisions of the AIA .
Information Disclosure Statement
The information disclosure statements (IDS) submitted on 3/27/2023, 7/26/2024, and 8/18/2025 are in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statements are being considered by the examiner.
Priority
Applicant claims the benefit of US Provisional Application No. 63/441,542, filed January 27, 2023. Claims 1-20 have been afforded the benefit of this filing date.
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 without significantly more.
Claims 1 and 11 recite A method for orchestrating an execution plan, the method comprising: receiving an input embedding, wherein the input embedding is generated by a machine- learning model; retrieving a plurality of stored semantic embeddings, from an embedding object memory, based on the input embedding, wherein the plurality of stored semantic embeddings each correspond to a respective historic plan, and wherein each historic plan comprises one or more executable skills; determining a subset of semantic embeddings from the plurality of stored semantic embeddings based on a similarity to the input embedding; generating, based on the subset of semantic embeddings and the input embedding, a new plan, wherein the new plan is different than the historic plans corresponding to the subset of semantic embeddings; and providing the new plan as an output.
The limitations in these claims, as drafted, are a process that, under broadest reasonable interpretation, covers performance of the limitation in the mind but for the recitation of generic computer components. The steps in the process are capable of being performed by the human mind. A human can receive an input embedding as this could just be a human language question that's been converted to a numerical representation by them. A human could then retrieve similar questions (in numerical form) from a journal they have them written in. These stored questions could all have a plan they’re associated with in the form of instructions written below the question. The skills in this sense are the individual actions performed in the instructions. The human would be capable of determining which entries in the journal are most similar to the input question using a formulaic approach. Then the human could take the instructions from the most similar journal entries and combine/alter them to form a new set of instructions for the input question. They could provide these instructions to whoever asked the input question by writing them down on a piece of paper. If a claim limitation, under its broadest reasonable interpretation, covers performance of the limitation in the mind but for the recitation of generic computer components, then it falls within the “Mental Processes” grouping of abstract ideas. Accordingly, the claim recites an abstract idea.
This judicial exception is not integrated into a practical application. The claims list the additional components of a machine-learning model. The machine-learning model is detailed in paragraph 74 of the specification with generic example models listed. Claim 11 specifically lists the additional components of a processor and memory. The processor is detailed in paragraph 93 of the specification with a generic description of the component. The memory is detailed in paragraph 95 of the specification with a generic description of the component. Accordingly, the claims do not include additional elements that are sufficient to amount to significantly more than the judicial exception. The claims are not patent eligible.
Claim 2 recites wherein the new plan comprises instructions that, when executed by a computing device, cause a set of operations to be performed corresponding to one or more skills.
The limitation in this claim, as drafted, is a process that, under broadest reasonable interpretation, covers performance of the limitation in the mind but for the recitation of generic computer components. A human can create a plan of instructions that execute on a computer by handwriting code in response to the input question. If a claim limitation, under its broadest reasonable interpretation, covers performance of the limitation in the mind but for the recitation of generic computer components, then it falls within the “Mental Processes” grouping of abstract ideas. Accordingly, the claim recites an abstract idea.
This judicial exception is not integrated into a practical application. The claim lists the additional components of a computing device. The computing device is detailed in paragraph 94 of the specification with generic description of the component. Accordingly, the claims do not include additional elements that are sufficient to amount to significantly more than the judicial exception. The claim is not patent eligible.
Claim 3 recites wherein the plurality of stored semantic embeddings each correspond to a respective historic input and the respective historic plan.
The limitation in this claim, as drafted, is a process that, under broadest reasonable interpretation, covers performance of the limitation in the mind but for the recitation of generic computer components. A human could keep a journal of past questions and the instructions provided to them. This could make up the journal they use to answer future questions. If a claim limitation, under its broadest reasonable interpretation, covers performance of the limitation in the mind but for the recitation of generic computer components, then it falls within the “Mental Processes” grouping of abstract ideas. Accordingly, the claim recites an abstract idea.
This judicial exception is not integrated into a practical application. The claim does not list any additional components that were not present in the independent claim. Accordingly, the claims do not include additional elements that are sufficient to amount to significantly more than the judicial exception. The claims are not patent eligible.
Claims 4 and 14 recite wherein the subset of semantic embeddings is further determined based on a personalization to at least one of a user or organization.
The limitation in these claims, as drafted, is a process that, under broadest reasonable interpretation, covers performance of the limitation in the mind but for the recitation of generic computer components. From the independent claim example, the selected journal entries could be further selected based on personal traits of the question asker as well as their similarity. If a claim limitation, under its broadest reasonable interpretation, covers performance of the limitation in the mind but for the recitation of generic computer components, then it falls within the “Mental Processes” grouping of abstract ideas. Accordingly, the claim recites an abstract idea.
This judicial exception is not integrated into a practical application. The claims do not list any additional components that were not present in the independent claim. Accordingly, the claims do not include additional elements that are sufficient to amount to significantly more than the judicial exception. The claims are not patent eligible.
Claims 5 and 15 recite wherein the personalization comprises: receiving metadata corresponding to the input embedding, wherein the subset of semantic embeddings are retrieved based on the similarity to the input embedding and the metadata.
The limitations in these claims, as drafted, are a process that, under broadest reasonable interpretation, covers performance of the limitation in the mind but for the recitation of generic computer components. From the independent claim example, the personal traits could be “metadata” such as a specific job the question asker has. They could match the input question to journal entries based on both similarity and the askers job. If a claim limitation, under its broadest reasonable interpretation, covers performance of the limitation in the mind but for the recitation of generic computer components, then it falls within the “Mental Processes” grouping of abstract ideas. Accordingly, the claim recites an abstract idea.
This judicial exception is not integrated into a practical application. The claims do not list any additional components that were not present in the independent claim. Accordingly, the claims do not include additional elements that are sufficient to amount to significantly more than the judicial exception. The claims are not patent eligible.
Claims 6 and 16 recite wherein the metadata is associated with compliance requirements for security.
The limitation in these claims, as drafted, is a process that, under broadest reasonable interpretation, covers performance of the limitation in the mind but for the recitation of generic computer components. From the independent claim example, the “metadata” could be compliance requirements for security. This could be based off background knowledge of the question asker. If a claim limitation, under its broadest reasonable interpretation, covers performance of the limitation in the mind but for the recitation of generic computer components, then it falls within the “Mental Processes” grouping of abstract ideas. Accordingly, the claim recites an abstract idea.
This judicial exception is not integrated into a practical application. The claims do not list any additional components that were not present in the independent claim. Accordingly, the claims do not include additional elements that are sufficient to amount to significantly more than the judicial exception. The claims are not patent eligible.
Claims 7 and 17 recite wherein the determining a subset of embeddings comprises: determining a respective similarity between the input embedding and each embedding of the plurality of stored semantic embeddings; determining an ordered ranking of the one or more similarities or that one or more of the similarities are less than a predetermined threshold; and identifying the subset of semantic embeddings with similarities to the input embedding that are less than the predetermined threshold or based on the ordered ranking, thereby retrieving a subset of semantic embeddings from the plurality of stored semantic embeddings that is determined to be related to the input embedding.
The limitations in these claims, as drafted, are a process that, under broadest reasonable interpretation, covers performance of the limitation in the mind but for the recitation of generic computer components. From the independent claim example, a human could choose a subset of journal entries that are similar to the input question and put them in a ranked order based on their similarity. They could also set a similarity threshold that is used to qualify journal entries to be in the set. Finally, they could use the ordered ranking and take entries that meet the threshold to help create an output. If a claim limitation, under its broadest reasonable interpretation, covers performance of the limitation in the mind but for the recitation of generic computer components, then it falls within the “Mental Processes” grouping of abstract ideas. Accordingly, the claim recites an abstract idea.
This judicial exception is not integrated into a practical application. The claims do not list any additional components that were not present in the independent claim. Accordingly, the claims do not include additional elements that are sufficient to amount to significantly more than the judicial exception. The claims are not patent eligible.
Claims 8 and 18 recite wherein the similarities are distances.
The limitation in these claims, as drafted, is a process that, under broadest reasonable interpretation, covers performance of the limitation in the mind but for the recitation of generic computer components. From the independent claim example, the similarities could be distances by using a formula on the numeric representation to calculate a similarity. If a claim limitation, under its broadest reasonable interpretation, covers performance of the limitation in the mind but for the recitation of generic computer components, then it falls within the “Mental Processes” grouping of abstract ideas. Accordingly, the claim recites an abstract idea.
This judicial exception is not integrated into a practical application. The claims do not list any additional components that were not present in the independent claim. Accordingly, the claims do not include additional elements that are sufficient to amount to significantly more than the judicial exception. The claims are not patent eligible.
Claims 9 and 19 recite wherein the input embedding and each embedding of the plurality of stored semantic embeddings are stored in a metric graph as nodes, wherein a respective edge is defined between the input embedding and each embedding of the plurality of stored semantic embeddings, and wherein each edge is associated with a respective distance of the distances.
The limitations in these claims, as drafted, are a process that, under broadest reasonable interpretation, covers performance of the limitation in the mind but for the recitation of generic computer components. A human is capable of creating a metric graph on a piece of paper. This could be used with the independent claim example where the input question and journal entries are written on the metric graph as nodes. Then the calculated distance could be written on the edges between the nodes. If a claim limitation, under its broadest reasonable interpretation, covers performance of the limitation in the mind but for the recitation of generic computer components, then it falls within the “Mental Processes” grouping of abstract ideas. Accordingly, the claim recites an abstract idea.
This judicial exception is not integrated into a practical application. The claims do not list any additional components that were not present in the independent claim. Accordingly, the claims do not include additional elements that are sufficient to amount to significantly more than the judicial exception. The claims are not patent eligible.
Claim 10 recites further comprising, prior to receiving the input embedding: receiving user-input; and generating the input embedding based on the user-input.
The limitations in this claim, as drafted, are a process that, under broadest reasonable interpretation, covers performance of the limitation in the mind but for the recitation of generic computer components. A human is capable of receiving an input question and converting it to a numerical representation. This could be done using a formula or defined rules. If a claim limitation, under its broadest reasonable interpretation, covers performance of the limitation in the mind but for the recitation of generic computer components, then it falls within the “Mental Processes” grouping of abstract ideas. Accordingly, the claim recites an abstract idea.
This judicial exception is not integrated into a practical application. The claim does not list any additional components that were not present in the independent claim. Accordingly, the claims do not include additional elements that are sufficient to amount to significantly more than the judicial exception. The claim is not patent eligible.
Claim 12 recites wherein the input is an input embedding, and wherein the input embedding is generated by a generative multimodal machine-learning model.
The limitations in this claim, as drafted, are a process that, under broadest reasonable interpretation, covers performance of the limitation in the mind but for the recitation of generic computer components. From the independent claim example, the input could be an embedding created by the human mind by converting natural language to a numerical representation using a formula or following predefined steps. If a claim limitation, under its broadest reasonable interpretation, covers performance of the limitation in the mind but for the recitation of generic computer components, then it falls within the “Mental Processes” grouping of abstract ideas. Accordingly, the claim recites an abstract idea.
This judicial exception is not integrated into a practical application. The claim lists the additional components of a machine-learning model. The machine-learning model is detailed in paragraph 74 of the specification with generic example models listed. Accordingly, the claims do not include additional elements that are sufficient to amount to significantly more than the judicial exception. The claim is not patent eligible.
Claim 13 recites wherein the set of operations further comprise: adapting a computing device to execute the plan.
The limitation in this claim, as drafted, is a process that, under broadest reasonable interpretation, covers performance of the limitation in the mind but for the recitation of generic computer components. A human can hand write a set of operations/code that a computer could be capable of executing. If a claim limitation, under its broadest reasonable interpretation, covers performance of the limitation in the mind but for the recitation of generic computer components, then it falls within the “Mental Processes” grouping of abstract ideas. Accordingly, the claim recites an abstract idea.
This judicial exception is not integrated into a practical application. The claim lists the additional components of a computing device. The computing device is detailed in paragraph 94 of the specification with generic description of the component. Accordingly, the claims do not include additional elements that are sufficient to amount to significantly more than the judicial exception. The claims are not patent eligible.
Claim 20 recites A method for orchestrating an execution plan, the method comprising: receiving an input; retrieving a plurality of stored semantic embeddings, from an embedding object memory, based on the input, wherein the plurality of stored semantic embeddings each correspond to a respective historic plan, and wherein each historic plan comprises one or more executable skills; determining a subset of semantic embeddings from the plurality of stored semantic embeddings based on a similarity to the input; filtering the subset of semantic embeddings based on a personalization associated with at least one of a user or organization, wherein the personalization is based on metadata associated with compliance requirements for security; generating, based on the filtered subset of semantic embeddings and the input, a new plan, wherein the new plan is different than the historic plans corresponding to the subset of semantic embeddings; and providing the new plan as an output.
The limitations in this claim, as drafted, are a process that, under broadest reasonable interpretation, covers performance of the limitation in the mind but for the recitation of generic computer components. A human can receive an input embedding as this could just be a human language question that's been converted to a numerical representation by them. A human could then retrieve similar questions (in numerical form) from a journal they have them written in. These stored questions could all have a plan they’re associated with in the form of instructions written below the question. The skills in this sense are the individual actions performed in the instructions. The human would be capable of determining which entries in the journal are most similar to the input question using a formulaic approach. They could also select the subset based on what they know about the question asker such as their occupation or clearance level for security purposes. Then the human could take the instructions from the most similar journal entries and combine/alter them to form a new set of instructions for the input question. They could provide these instructions to whoever asked the input question by writing them down on a piece of paper. If a claim limitation, under its broadest reasonable interpretation, covers performance of the limitation in the mind but for the recitation of generic computer components, then it falls within the “Mental Processes” grouping of abstract ideas. Accordingly, the claim recites an abstract idea.
This judicial exception is not integrated into a practical application. The claim lists the additional components of a memory. The memory is detailed in paragraph 95 of the specification with a generic description of the component. Accordingly, the claims do not include additional elements that are sufficient to amount to significantly more than the judicial exception. The claims are 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.
Claims 1-3, 7-8, 10-13, and 17-18 are rejected under 35 U.S.C. 103 as being unpatentable over US Patent Publication US 12079185 B2 (Misiewicz et al.) in view of US Patent Publication US 11720346 B2 (Wu et al.).
Regarding Claims 1 and 11, Misiewicz et al. teaches A method for orchestrating an execution plan, the method comprising:
(A system and method to generate search results in response to a search query based on comparisons of embedding vectors. The system and method receive, from an end user system, a search query including a set of keywords associated with the entity.) (Abstract).
Claim 11 lists the alternative limitation of: A system for orchestrating an execution plan, the system comprising: a processor; and memory storing instructions that, when executed by the processor, cause the system to perform a set of operations, the set of operations comprising:
(The example computer system 500 may comprise a processing device 502 (also referred to as a processor or CPU), a main memory 504 (e.g., read-only memory (ROM), flash memory, dynamic random access memory (DRAM) such as synchronous DRAM (SDRAM), etc.), a static memory 506 (e.g., flash memory, static random access memory (SRAM), etc.), and a secondary memory (e.g., a data storage device 516), which may communicate with each other via a bus 530.) (Col. 13, Line 66 to Col. 14, Line 7).
receiving an input embedding, wherein the input embedding is generated by a machine- learning model;
(In operation 320, the processing logic identifies, using a neural network, an embedding vector based on the set of keywords of the search query. In an embodiment, a request for an embedding vector representing the search query is sent to the neural network.) (Col. 11, Lines 24-28).
(According to an embodiment, the embedding vectors for each structured data elements can be generated using a neural network or machine learning system configured to employ a library for learning term embeddings and term classifications to create an unsupervised or supervised learning algorithm to generate the vector representations for the search terms.) (Col. 5, Lines 12-18).
In Misiewicz et al. the user submits an initial search query which is then made into an embedding vector by a neural network or MLM.
retrieving a plurality of stored semantic embeddings, from an embedding object memory, based on the input embedding,
(In operation 330, the processing logic executes a comparison of the embedding vector associated with the search query to a set of embedding vectors associated with a set of structured data elements relating to the entity. In an embodiment, the set of embedding vectors associated with the set of structured data elements is generated and stored in an indexed data store, as described in method 200.) (Col. 11, Lines 34-40).
The input embedding is compared to a set of stored embeddings related to the same entity.
wherein the plurality of stored semantic embeddings each correspond to a respective historic plan,
(According to embodiments of the present disclosure, the knowledge search system is configured to efficiently identify search results including FAQs that are responsive to a keyword-based search query. In an embodiment, the knowledge search system builds and maintains indices of the FAQs (e.g., structured data elements in the form of frequently asked questions that can be returned as a search result) associated with a particular entity. In each generated entity-specific index, each FAQ is associated a “name” field and a generated embedding vector associated with the FAQ.) (Col. 12, Lines 41-51).
The structured data elements represent the semantic embeddings. These elements are frequently asked questions (FAQs) that relate to the search query. This represents a historic plan where the question part of the FAQ is a historic query and the corresponding answer is the historic plan to respond to that query.
determining a subset of semantic embeddings from the plurality of stored semantic embeddings based on a similarity to the input embedding;
(In operation 340, the processing logic identifies, based on the comparison, a set of matching structured data elements. In an embodiment, the processing logic applies a distance threshold to each of the generated distance metric function scores. In an embodiment, the set of matching structured data elements is identified as the structured data element that have a distance-comparison score that is less than the distance threshold level, wherein smaller distance results represent structured data elements that are determined to be more relevant in response to the search query.) (Col. 11, Lines 56-65).
Through the similarity analysis, a subset of the structured data elements (FAQs) is determined as the elements that are below a distance threshold.
generating, based on the subset of semantic embeddings and the input embedding, a new plan,
(In operation 350, the processing logic generate a search result in response to the search query, wherein the search result includes at least a portion of the set of matching structured data elements.) (Col. 12, Lines 8-11).
(For example, if the search query is “New York City”, … a “FAQ” search provider source can provide a set of search responses including frequently asked questions associated with New York City.) (Col. 8, Lines 49-57).
The response to the search query represents a new plan that is constructed from the similar FAQ response.
wherein the new plan is different than the historic plans corresponding to the subset of semantic embeddings;
(In operation 350, the processing logic generate a search result in response to the search query, wherein the search result includes at least a portion of the set of matching structured data elements. In an embodiment, the set of matching structured data elements (e.g., the elements having a distance in the numerical vector space that is less than the distance threshold) can be further ranked based on their relative distance scores…. For example, a “top” or highest priority position in the interface can be assigned to the matching structured data element having a lowest relative distance score, a second highest priority position in the interface can be can be assigned to the matching structured data element having a second lowest relative distance score, and so on.) (Col. 12, Lines 8-37)
A response (new plan) is created based on the FAQs in the structured data elements (historic plan). These elements can be included/excluded and reordered based on their relevance to the input question. The combination of the historic plans creates a new plan for the user.
and providing the new plan as an output.
(In operation 360, the processing logic causes a display of an interface via the end user system including the search result.) (Col 12, Lines 25-27).
The result is output to the user
Misiewicz et al. does not explicitly teach: and wherein each historic plan comprises one or more executable skills;
However, Wu et al. teaches and wherein each historic plan comprises one or more executable skills;
(At 1106, the computer-implemented method 1100 can comprise generating (e.g., via graph construction component 114), by the system 100, a graph representation of at least a portion of computer program code (e.g., a code snippet) from one or more source code repositories 108.) (Col. 18, Lines 16-20).
(Given the problems with other implementations of code retrieval tasks; the present disclosure can be implemented to produce a solution to one or more of these problems by employing an end-to-end graph-based model that can exploit rich structural information in both natural language query texts and code repositories.) (Col. 3, Line 66 to Col. 4, Line 4).
Wu et al. deploys a similar graph-based similarity search system; however, Wu et al. finds sections of code from a repository. By finding code the historic plans in this instance are executables that could be considered skills.
It would have been obvious to a person having ordinary skill in the art before the time of the effective filing date of the claimed invention of the instant application to modify the graph-based similarity search method as taught by Misiewicz et al. to retrieve code segments instead of FAQs related to the initial input as taught by Wu et al. This would have been an obvious substitutions as it is a similar method only it is being applied to a different source (code repository vs. FAQ searches). Furthermore, there is motivation to apply this method to code repositories as it can accelerate the time it takes for users to find code snippets among billions of lines of source code (Wu et al. Col. 1, Lines 12-25).
Regarding Claim 2, Misiewicz et al. in view of Wu et al. teaches the system of claim 1.
Furthermore, Wu et al. teaches wherein the new plan comprises instructions that, when executed by a computing device, cause a set of operations to be performed corresponding to one or more skills.
(Also, one or more embodiments described herein can constitute a technical improvement over conventional code retrieval task approaches by exploiting graph structure information of the natural language query text and the source code snippets included in the code repositories for improving code matching accuracy and ranking performance. Additionally, various embodiments described herein can demonstrate a technical improvement over conventional code retrieval task approaches by employing a graph-based approach to overcome semantic differences between the structure of the natural language query text and the one or more code snippets.) (Col. 4, Lines 53-64)
Wu et al. uses a similar graph based semantic similarity method to retrieve code-snippets based on an input. In this case the new plan is the output which is the code snippet. By definition, a code-snippet causes a set of operations corresponding to a skill when executed by a computer.
Regarding Claim 3, Misiewicz et al. in view of Wu et al. teaches the system of claim 1.
Furthermore, Misiewicz et al. teaches wherein the plurality of stored semantic embeddings each correspond to a respective historic input and the respective historic plan.
(In an embodiment, the knowledge search system builds and maintains indices of the FAQs (e.g., structured data elements in the form of frequently asked questions that can be returned as a search result) associated with a particular entity.) (Col. 12, Lines 41-51)
(The knowledge search system is configured to simulate vector search using historical or actual queries and the FAQs and rate the query/FAQ pairs that were returned by an un-trained neural network model. In this embodiment, the FAQs that were returned using the simulation (e.g., the set of FAQs that would have been returned using an untrained vector search) to generate training data to be used as the basis for collecting labeled data.) (Col. 14, Lines 60-67).
The stored embeddings correspond to historic input in the sense that they are also queries that have been asked before. Furthermore, the system uses historical queries to further improve the vector search system.
Regarding Claims 7 and 17, Misiewicz et al. in view of Wu et al. teaches the system of claims 1 and 11.
Furthermore, Misiewicz et al. teaches wherein the determining a subset of embeddings comprises: determining a respective similarity between the input embedding and each embedding of the plurality of stored semantic embeddings;
(In operation 330, the processing logic executes a comparison of the embedding vector associated with the search query to a set of embedding vectors associated with a set of structured data elements relating to the entity. In an embodiment, the set of embedding vectors associated with the set of structured data elements is generated and stored in an indexed data store, as described in method 200. In an embodiment, the comparison includes a matching determination based on a relative similarity in a numerical vector space between the respective embedding vectors of the set of structured data elements and the embedding vector associated with the search query.) (Col. 11, Lines 34-45).
The system finds a measured similarity between the embedded input and the stored embeddings.
determining an ordered ranking of the one or more similarities or that one or more of the similarities are less than a predetermined threshold;
(In operation 340, the processing logic identifies, based on the comparison, a set of matching structured data elements. In an embodiment, the processing logic applies a distance threshold to each of the generated distance metric function scores.) (Col. 11, Lines 56-60).
(For example, a first matching structured data element having a lowest relative distance score (e.g., being the closest to the embedding vector of the search query in the numerical vector space) can have a highest ranking, a second matching structured data element having a second lowest relative distance score can have a second highest ranking, and so on. In an embodiment, at least a portion of the matching structured data elements can be selected for inclusion in the search result.) (Col. 12, Lines 16-24)
Vector distances are used to find similarities between the input and the set of data elements. A distance threshold can be used to create a subset of the set for the output. Furthermore, the distance measurements can be used to rank the set of data elements and the set can be narrowed that way.
and identifying the subset of semantic embeddings with similarities to the input embedding that are less than the predetermined threshold or based on the ordered ranking,
(In an embodiment, the set of matching structured data elements is identified as the structured data element that have a distance-comparison score that is less than the distance threshold level, wherein smaller distance results represent structured data elements that are determined to be more relevant in response to the search query.) (Col. 11, Lines 60-65).
(In an embodiment, as an alternative to applying the distance threshold to identify the matching structured data elements, the processing logic can employ a top-N approach, wherein an N number of structured data elements are identified as matching.) (Col. 12, Lines 3-7).
The set can be selected based on a distance threshold or by selecting a set number based on the ordered ranking from the distance values.
thereby retrieving a subset of semantic embeddings from the plurality of stored semantic embeddings that is determined to be related to the input embedding.
(In operation 350, the processing logic generate a search result in response to the search query, wherein the search result includes at least a portion of the set of matching structured data elements.) (Col. 12, Lines 8-12).
The set of matching elements is retrieved and used to form the response.
Regarding Claims 8 and 18, Misiewicz et al. in view of Wu et al. teaches the system of claims 7 and 17.
Furthermore, Misiewicz et al. teaches wherein the similarities are distances.
(In an embodiment, the comparison includes scoring each of the structured data elements based on a distance metric function (e.g., a cosine similarity function, an Euclidean distance function, a Manhattan distance function, a Jaccard Similarity function, etc.).) (Col. 11, Lines 50-55).
The similarity measurements are done using distance.
Regarding Claim 10, Misiewicz et al. in view of Wu et al. teaches the system of claim 1.
Furthermore, Misiewicz et al. teaches further comprising, prior to receiving the input embedding: receiving user-input;
(In operation 310, the processing logic receives, from an end user system, a search query including a set of keywords associated with an entity.) (Col. 11, Lines 19-21).
The user submits an input in the form of an initial search query.
and generating the input embedding based on the user-input.
(In operation 320, the processing logic identifies, using a neural network, an embedding vector based on the set of keywords of the search query. In an embodiment, a request for an embedding vector representing the search query is sent to the neural network.) (Col. 11, Lines 24-28).
The users initial search query is then made into an embedding vector by a neural network or MLM.
Regarding Claim 12, Misiewicz et al. in view of Wu et al. teaches the system of claim 11.
Furthermore, Misiewicz et al. teaches wherein the input is an input embedding,
(In operation 320, the processing logic identifies, using a neural network, an embedding vector based on the set of keywords of the search query. In an embodiment, a request for an embedding vector representing the search query is sent to the neural network.) (Col. 11, Lines 24-28).
In Misiewicz et al. the user submits an initial search query which is then made into an embedding vector that the system uses in the following steps.
and wherein the input embedding is generated by a generative multimodal machine-learning model.
(In operation 320, the processing logic identifies, using a neural network, an embedding vector based on the set of keywords of the search query. In an embodiment, a request for an embedding vector representing the search query is sent to the neural network.) (Col. 11, Lines 24-28).
(According to an embodiment, the embedding vectors for each structured data elements can be generated using a neural network or machine learning system configured to employ a library for learning term embeddings and term classifications to create an unsupervised or supervised learning algorithm to generate the vector representations for the search terms.) (Col. 5, Lines 12-18).
The users initial query is made into an embedding vector by a neural network or MLM.
Regarding Claim 13, Misiewicz et al. in view of Wu et al. teaches the system of claim 11.
Furthermore, Misiewicz et al. teaches wherein the set of operations further comprise: adapting a computing device to execute the plan.
(The knowledge search system 110 is configured to receive the search query 40 and generate a set of search results 180 in response to the search query 40 initiated by the end user system 10. For example, the end user system 10 may be any suitable computing device (e.g., a mobile device, a desktop computer, a laptop computer, etc.) associated with an end user in search of information relating to the entity (e.g., information about a merchant or a related product or service).) (Col. 4, Lines 26-34).
The query and resulting response are received/given on an end user system that is a computing device.
Claims 4-6, 14-16, and 20 are rejected under 35 U.S.C. 103 as being unpatentable over US Patent Publication US 12079185 B2 (Misiewicz et al.) in view of US Patent Publication US 11720346 B2 (Wu et al.) and further in view of US Patent Publication US 11886608 B2 (Hansen et al.).
Regarding Claims 4 and 14, Misiewicz et al. in view of Wu et al. teaches the system of claims 1 and 11.
Misiewicz et al. in view of Wu et al. does not explicitly teach: wherein the subset of semantic embeddings is further determined based on a personalization to at least one of a user or organization.
However, Hansen et al. teaches wherein the subset of semantic embeddings is further determined based on a personalization to at least one of a user or organization.
(In some embodiments, in response to execution of a task (e.g., running a task against a dataset in response to a user request to execute the task, performing a query against the log of user activity, etc.), database system 105 determines information responsive to execution of the task, and security system 135 determines a subset of the information responsive to execution of the task for which the user associated with the request to execute the task/query has the requisite access permissions. In some embodiments, security system 135 filters information and outputs only information that the requesting user is permitted to access.) (Col. 9, Lines 14-24).
Hansen et al. presents a method that further filters the information that will be given to the user based on the user’s security permissions. Security permissions are considered personalization for the user or organization.
It would have been obvious to a person having ordinary skill in the art before the time of the effective filing date of the claimed invention of the instant application to modify the graph-based similarity search method as taught by Misiewicz et al. in view of Wu et al. to consider personal information about the user in order to restrict the information output to them as taught by Hansen et al. This would have been an obvious improvement as large databases may contain information that not all users or organizations have access to (Hansen et al. Col. 9, Lines 1-13).
Regarding Claims 5 and 15, Misiewicz et al. in view of Wu et al. and Hansen et al. teaches the system of claims 4 and 14.
Furthermore, Hansen et al. teaches wherein the personalization comprises: receiving metadata corresponding to the input embedding,
(System 200 uses information access module 240 to determine at least a subset of information that is responsive to the execution of the task that the user is permitted to access (e.g., information that is permitted to be returned to the user). Information access module 240 is implemented by database system 105 of system 100 of FIG. 1, and/or security system 135. In some embodiments, information access module 240 provides to a security system the information that is responsive to execution to the task, and receives the subset of such information that the user is permitted to access. For example, the security system filters the responsive information to include only such information for which the user has the applicable access permissions.) (Col. 11, Lines 34-46).
The system considers the task the user is trying to execute (input embedding) and filters information according to the security information of the user (metadata).
wherein the subset of semantic embeddings are retrieved based on the similarity to the input embedding and the metadata.
(The one or more processors are configured to receive a request to execute a task with respect to a database, wherein the request is associated with an identifier corresponding to a user that inputs a query for the request; determine whether the task is authorized for the user; in response to a determination that the task is authorized for the user, obtain a set of information that is to be returned for the task; determine a subset of the set of information, wherein the subset of the set of information comprises one or more parts of the set of information for which the user has access permission;) (Col. 9, Lines 14-24).
A set of information is retrieved based on the users’ specific input, that set is then further narrowed by the metadata in the form of security permissions.
Regarding Claims 6 and 16, Misiewicz et al. in view of Wu et al. and Hansen et al. teaches the system of claims 5 and 15.
Furthermore, Hansen et al. teaches wherein the metadata is associated with compliance requirements for security.
(In some embodiments, in response to execution of a task (e.g., running a task against a dataset in response to a user request to execute the task, performing a query against the log of user activity, etc.), database system 105 determines information responsive to execution of the task, and security system 135 determines a subset of the information responsive to execution of the task for which the user associated with the request to execute the task/query has the requisite access permissions. In some embodiments, security system 135 filters information and outputs only information that the requesting user is permitted to access.) (Col. 9, Lines 14-24).
The metadata is security compliance information.
Regarding Claim 20, Misiewicz et al. teaches A method for orchestrating an execution plan, the method comprising:
(A system and method to generate search results in response to a search query based on comparisons of embedding vectors. The system and method receive, from an end user system, a search query including a set of keywords associated with the entity.) (Abstract).
receiving an input;
(In operation 320, the processing logic identifies, using a neural network, an embedding vector based on the set of keywords of the search query. In an embodiment, a request for an embedding vector representing the search query is sent to the neural network.) (Col. 11, Lines 24-28).
(According to an embodiment, the embedding vectors for each structured data elements can be generated using a neural network or machine learning system configured to employ a library for learning term embeddings and term classifications to create an unsupervised or supervised learning algorithm to generate the vector representations for the search terms.) (Col. 5, Lines 12-18).
In Misiewicz et al. the user submits an initial search query which is then made into an embedding vector by a neural network or MLM.
retrieving a plurality of stored semantic embeddings, from an embedding object memory, based on the input,
(In operation 330, the processing logic executes a comparison of the embedding vector associated with the search query to a set of embedding vectors associated with a set of structured data elements relating to the entity. In an embodiment, the set of embedding vectors associated with the set of structured data elements is generated and stored in an indexed data store, as described in method 200.) (Col. 11, Lines 34-40).
The input embedding is compared to a set of stored embeddings related to the same entity.
wherein the plurality of stored semantic embeddings each correspond to a respective historic plan,
(According to embodiments of the present disclosure, the knowledge search system is configured to efficiently identify search results including FAQs that are responsive to a keyword-based search query. In an embodiment, the knowledge search system builds and maintains indices of the FAQs (e.g., structured data elements in the form of frequently asked questions that can be returned as a search result) associated with a particular entity. In each generated entity-specific index, each FAQ is associated a “name” field and a generated embedding vector associated with the FAQ.) (Col. 12, Lines 41-51).
The structured data elements represent the semantic embeddings. These elements are frequently asked questions (FAQs) that relate to the search query. This represents a historic plan where the question part of the FAQ is a historic query and the corresponding answer is the historic plan to respond to that query.
determining a subset of semantic embeddings from the plurality of stored semantic embeddings based on a similarity to the input;
(In operation 340, the processing logic identifies, based on the comparison, a set of matching structured data elements. In an embodiment, the processing logic applies a distance threshold to each of the generated distance metric function scores. In an embodiment, the set of matching structured data elements is identified as the structured data element that have a distance-comparison score that is less than the distance threshold level, wherein smaller distance results represent structured data elements that are determined to be more relevant in response to the search query.) (Col. 11, Lines 56-65).
Through the similarity analysis, a subset of the structured data elements (FAQs) is determined as the elements that are below a distance threshold.
generating, based on the filtered subset of semantic embeddings and the input, a new plan,
(In operation 350, the processing logic generate a search result in response to the search query, wherein the search result includes at least a portion of the set of matching structured data elements.) (Col. 12, Lines 8-11).
(For example, if the search query is “New York City”, … a “FAQ” search provider source can provide a set of search responses including frequently asked questions associated with New York City.) (Col. 8, Lines 49-57).
The response to the search query represents a new plan that is constructed from the similar FAQ response.
wherein the new plan is different than the historic plans corresponding to the subset of semantic embeddings;
(In operation 350, the processing logic generate a search result in response to the search query, wherein the search result includes at least a portion of the set of matching structured data elements. In an embodiment, the set of matching structured data elements (e.g., the elements having a distance in the numerical vector space that is less than the distance threshold) can be further ranked based on their relative distance scores…. For example, a “top” or highest priority position in the interface can be assigned to the matching structured data element having a lowest relative distance score, a second highest priority position in the interface can be can be assigned to the matching structured data element having a second lowest relative distance score, and so on.) (Col. 12, Lines 8-37)
A response (new plan) is created based on the FAQs in the structured data elements (historic plan). These elements can be included/excluded and reordered based on their relevance to the input question. The combination of the historic plans creates a new plan for the user.
and providing the new plan as an output.
(In operation 360, the processing logic causes a display of an interface via the end user system including the search result.) (Col 12, Lines 25-27).
The result is output to the user.
Misiewicz et al. does not explicitly teach: and wherein each historic plan comprises one or more executable skills; filtering the subset of semantic embeddings based on a personalization associated with at least one of a user or organization, wherein the personalization is based on metadata associated with compliance requirements for security;
However, Wu et al. teaches and wherein each historic plan comprises one or more executable skills;
(At 1106, the computer-implemented method 1100 can comprise generating (e.g., via graph construction component 114), by the system 100, a graph representation of at least a portion of computer program code (e.g., a code snippet) from one or more source code repositories 108.) (Col. 18, Lines 16-20).
(Given the problems with other implementations of code retrieval tasks; the present disclosure can be implemented to produce a solution to one or more of these problems by employing an end-to-end graph-based model that can exploit rich structural information in both natural language query texts and code repositories.) (Col. 3, Line 66 to Col. 4, Line 4).
Wu et al. deploys a similar graph-based similarity search system; however, Wu et al. finds sections of code from a repository. By finding code the historic plans in this instance are executables that could be considered skills.
It would have been obvious to a person having ordinary skill in the art before the time of the effective filing date of the claimed invention of the instant application to modify the graph-based similarity search method as taught by Misiewicz et al. to retrieve code segments instead of FAQs related to the initial input as taught by Wu et al. This would have been an obvious substitutions as it is a similar method only it is being applied to a different source (code repository vs. FAQ searches). Furthermore, there is motivation to apply this method to code repositories as it can accelerate the time it takes for users to find code snippets among billions of lines of source code (Wu et al. Col. 1, Lines 12-25).
Misiewicz et al. in view of Wu et al. does not explicitly teach: filtering the subset of semantic embeddings based on a personalization associated with at least one of a user or organization, wherein the personalization is based on metadata associated with compliance requirements for security;
However, Hansen et al. teaches filtering the subset of semantic embeddings based on a personalization associated with at least one of a user or organization,
(In some embodiments, in response to execution of a task (e.g., running a task against a dataset in response to a user request to execute the task, performing a query against the log of user activity, etc.), database system 105 determines information responsive to execution of the task, and security system 135 determines a subset of the information responsive to execution of the task for which the user associated with the request to execute the task/query has the requisite access permissions. In some embodiments, security system 135 filters information and outputs only information that the requesting user is permitted to access.) (Col. 9, Lines 14-24).
Hansen et al. presents a method that further filters the information that will be given to the user based on the user’s security permissions. Security permissions are considered personalization for the user or organization.
wherein the personalization is based on metadata associated with compliance requirements for security;
(In some embodiments, in response to execution of a task (e.g., running a task against a dataset in response to a user request to execute the task, performing a query against the log of user activity, etc.), database system 105 determines information responsive to execution of the task, and security system 135 determines a subset of the information responsive to execution of the task for which the user associated with the request to execute the task/query has the requisite access permissions. In some embodiments, security system 135 filters information and outputs only information that the requesting user is permitted to access.) (Col. 9, Lines 14-24).
The metadata is security compliance information.
It would have been obvious to a person having ordinary skill in the art before the time of the effective filing date of the claimed invention of the instant application to modify the graph-based similarity search method as taught by Misiewicz et al. in view of Wu et al. to consider personal information about the user in order to restrict the information output to them as taught by Hansen et al. This would have been an obvious improvement as large databases may contain information that not all users or organizations have access to (Hansen et al. Col. 9, Lines 1-13).
Claims 9 and 19 are rejected under 35 U.S.C. 103 as being unpatentable over US Patent Publication US 12079185 B2 (Misiewicz et al.) in view of US Patent Publication US 11720346 B2 (Wu et al.) and further in view of US Patent Publication US 12169512 B2 (Hamilton et al.).
Regarding Claims 9 and 19, Misiewicz et al. in view of Wu et al. teaches the system of claims 8 and 18.
Misiewicz et al. in view of Wu et al. does not explicitly teach: wherein the input embedding and each embedding of the plurality of stored semantic embeddings are stored in a metric graph as nodes, wherein a respective edge is defined between the input embedding and each embedding of the plurality of stored semantic embeddings, and wherein each edge is associated with a respective distance of the distances.
However, Hamilton et al. teaches wherein the input embedding and each embedding of the plurality of stored semantic embeddings are stored in a metric graph as nodes,
(a query input embedding of the query input is generated with respect to the content category; for each query input embedding, a k-Nearest-Neighbor (KNN) search is performed with respect to search engine repository item embeddings to generate initial search results; for each initial set result, performing N hops within a semantic graph starting from nodes associated with the initial search result to generate related search results;) (Col. 1, Lines 32-39).
(The term “semantic graph” may refer to a data construct that describes a graph including nodes that are representative of search engine repository items from a search engine repository and edges representative of similarity between semantic embeddings of the search engine repository items.) (Col. 9, Lines 17- 21).
Hamilton et al. presents a similar graph-based semantic similarity method which specifically states storing the embeddings on a semantic graph.
wherein a respective edge is defined between the input embedding and each embedding of the plurality of stored semantic embeddings,
(The term “semantic graph” may refer to a data construct that describes a graph including nodes that are representative of search engine repository items from a search engine repository and edges representative of similarity between semantic embeddings of the search engine repository items.) (Col. 9, Lines 17- 21).
Edges are drawn between the nodes representing their similarity to each other.
and wherein each edge is associated with a respective distance of the distances.
(The term “k-Nearest-Neighbor (KNN) search” may refer to a search technique for finding K nearest vectors to a query input embedding vector according to a similarity metric, such as Euclidean distance or cosine similarity. According to various embodiments of the present disclosure, a KNN search may be performed between query input embeddings of a query input and search engine repository item embeddings of search engine repository items to retrieve top K candidate search engine repository items along with respective cosine similarity scores.) (Col. 8, Lines 32-41).
The similarities that the edges represent are calculated Euclidian distances or cosine similarities representative of the distance.
It would have been obvious to a person having ordinary skill in the art before the time of the effective filing date of the claimed invention of the instant application to modify the graph-based similarity search method as taught by Misiewicz et al. in view of Wu et al. to include the graph with the embeddings as nodes and corresponding edges as taught by Hamilton et al. This would have been an obvious improvement as Misiewicz et al. is already calculating these distances in their method. Furthermore, the method of Hamilton et al. is stated to help improve the quality of search results which is also the goal of Misiewicz et al. (Hamilton et al. Col. 1, Lines 15-19).
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to NICHOLAS DANIEL LOWEN whose telephone number is (571)272-5828. The examiner can normally be reached Mon-Fri 8:00am - 4:00pm.
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.
/NICHOLAS D LOWEN/Examiner, Art Unit 2653
/DANIEL C WASHBURN/Supervisory Patent Examiner, Art Unit 2657