DETAILED ACTION
This communication is a Final Action in response to correspondence filed on July 2, 2026. Claims 1 and 10 have been amended. Claims 2 and 11 have been cancelled. Claims 1, 3-10, and 12-18 are pending in the application.
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 .
Drawings
The drawings are objected to as failing to comply with 37 CFR 1.84(p)(5) because they include the following reference character(s) not mentioned in the description: FIG. 3B. Corrected drawing sheets in compliance with 37 CFR 1.121(d), or amendment to the specification to add the reference character(s) in the description in compliance with 37 CFR 1.121(b) are required in reply to the Office action to avoid abandonment of the application. Any amended replacement drawing sheet should include all of the figures appearing on the immediate prior version of the sheet, even if only one figure is being amended. Each drawing sheet submitted after the filing date of an application must be labeled in the top margin as either “Replacement Sheet” or “New Sheet” pursuant to 37 CFR 1.121(d). If the changes are not accepted by the examiner, the applicant will be notified and informed of any required corrective action in the next Office action. The objection to the drawings will not be held in abeyance.
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.
This application currently names joint inventors. In considering patentability of the claims the examiner presumes that the subject matter of the various claims was commonly owned as of the effective filing date of the claimed invention(s) absent any evidence to the contrary. Applicant is advised of the obligation under 37 CFR 1.56 to point out the inventor and effective filing dates of each claim that was not commonly owned as of the effective filing date of the later invention in order for the examiner to consider the applicability of 35 U.S.C. 102(b)(2)(C) for any potential 35 U.S.C. 102(a)(2) prior art against the later invention.
Claims 1, 3-10, and 12-18 are rejected under 35 U.S.C. 103 as being unpatentable over Bansal et al. (US Publication No. 2021/0004770) and Zhao (US Publication No. 2026/0080347) in view of Guerra et al. (US Patent No. 12,346,314) and further in view of Farre Guiu et al. (US Publication No. 2017/0109419).
As to claim 1, Bansal teaches a method comprising:
retrieving, at a server [server 106], metadata [event features/attributes] of one or more existing database entities [event data objects] (see e.g., [0034] for server 106 being provided via multiple devices arranged in a distributed environment that collectively provide the functionality described herein, [0035] for server 106 comprising server-side software designed to work in conjunction with client-side software on user devices 102a and 102b through 102n so as to implement any combination of the features and functionalities discussed in the present disclosure, [0061] for event content extractor 266 then extracting event content from information about the user-related activity and event content generally referring to any information associated with an event, including but not limited to information about a service provider, location, name of event, date, event attendees, confirmation or reservation information, tracking or itinerary numbers, information specific to event types such as flight numbers, check-in/pick-up and check-out/drop-off times, agendas, or other details that are specific to a particular event, [0063] for event content extractor 266 also generating an indication or record of the event, which may comprise an event data object, and which may be stored in event data objects (entities) 242, a user profile 240, or otherwise stored and associated with the user, [0064] for event content enricher 268, in general, enriching event data objects, or the event content, with supplemental information that is relevant to that particular event and supplemental information including contextual information for the event that is not expressly included in a particular instance of user-related activity, for example and without limitation, supplemental information including: enriched location-related content (e.g., determining geo-location from extracted content, such as determining city, region, latitude-longitude from airport codes for travel); contact information and other information about service providers associated with an event (e.g., a hotel's phone number, email, or website); or nearly any other information related to the event, and extracted content corresponding to an event, or aspects of the user-related activity information associated with the event being utilized by event content enricher 268 to generate supplemental information, [0065] for new event integrator 270 utilizing the supplemental information for determining whether an event is new or related to other events, [0074] for new event integrator 270 comprising a new event identifier 272, new event identifier 272 receiving information about an event, such as an indication of the event or an event data object for the event, and analyzing the information to determine whether the corresponding event is new, such as by comparing the information to other information about the other, previously detected events associated with the user, and information about the previously detected events being determined or received from event data objects corresponding to those events, which may be stored in a user profile 240 associated with the user, such as in event data objects (entities) 242, and [0075] for new event identifier 272 performing the analysis to determine whether a detected event is new by comparing a set of features or attributes of the newly detected event to features or attributes of previously detected events associated with the user. Server 106, which may implement the new event integrator 270, retrieves event features/attributes of an existing event data object. The event features/attributes include context information corresponding to the event data object.);
receiving, at the server, data model attributes for a new database entity and generating new input data based on the data model attributes (see e.g., [0025] for the event content being identified utilizing data models, [0116] for the term “data object” or “entity” being used broadly herein and comprising a record or set of records, data organized according to a data model (which may include a set of one or more data objects), a data structure, or template (which may be pre-defined and/or may be based on the particular category of event), a class object, a combination of these, or similar data structure and for instance, in some embodiments, the data structure comprising a pre-defined structure for each event type or being defined from a data structure template or data model, [0063] for event content extractor 266 generating an event data object (or an event entity) corresponding to a particular event that comprises the extracted event content from user-related activity information indicating or associated with the event and in some embodiments, event content extractor 266 providing the event data object, or extracted event content, to event content enricher 268, [0064] for event content enricher 268, in general, enriching event data objects, or the event content, with supplemental information that is relevant to that particular event, and [0074] for new event identifier 272 receiving information about an event, such as an indication of the event or an event data object for the event, and analyzing the information to determine whether the corresponding event is new, such as by comparing the information to other information about the other, previously detected events associated with the user. The server, which implements event detector 260 and new event integrator 270, receives data model attributes for the purpose of providing an organized structure for new event data objects. New event content extraction data and event content enrichment data are generated based on the model attributes. These new event features/attributes are received by the new event identifier for comparison purposes.); and
receiving, at the sever, a recommendation of whether at least one of the existing database entities that is determined to be similar to the new database entity is to be added to or modified based on the generated new input metadata (see e.g., [0075] for new event identifier 272 performing the analysis to determine whether a detected event is new by comparing a set of features or attributes of the newly detected event to features or attributes of previously detected events associated with the user, where the similarity score satisfies a threshold, this implying the newly detected event is sufficiently similar to a previously detected event, and thus is not new, and new event identifier 272 utilizing the enriched information from event content enricher 268 to increase the accuracy of identifying whether an event is new, [0099] for new event integrator 270 determining that an event data object does not correspond to a new event, and updated event determiner 282 determining that an event comprises an update and which event data object(s) should be updated, [0101] comparing an update entity with existing event entities of the same event type, and then determining whether attributes in the update entity match attributes in the existing event entities of that type, in order to determine the specific existing entity for which the update regards, the attributes or features used for the comparison comprising extracted or supplemental event content, and where the similarity score satisfies a threshold, the one or more entities being determined to be associated with the update entity, and [0102] for event entity reconciliation 284 being generally responsible for reconciling information from an update entity and information from existing event entities for which the update regards, the content in the existing event entity being modified based on the content from the update entity, and the modification comprising changing existing attributes in the existing event entity or adding to the existing event entity additional or new information from the update entity. The server, which implements event entity reconciliation 286, receives a recommendation of whether at least one of the existing event data objects that is determined to be similar to the new event data object is to be added to or modified based on the newly generated event content extraction data and event content enrichment data.).
Bansal does not specifically disclose generating vector embeddings for the one or more existing database entities; generating, at the server, a new vector embedding based on the generated new input metadata; determining, at the server, one or more similarities between the generated new vector embedding and at least one of the generated vector embeddings of the one or more existing database entities; at least one of the existing database entities being determined to be similar to the new database entity when at least one of the corresponding generated vector embeddings is determined to be similar to the corresponding new vector embedding; and a recommendation to modify at least one of the existing database entities constituting a recommendation to modify at least one of the corresponding generated vector embeddings. However, Zhao teaches
generating vector embeddings for the one or more existing database entities [sub-structures/chunks/sub-hierarchies] (see e.g., [0030] for the auto-assignment techniques including dividing the hierarchical structure into smaller sub-structures, or relevant chunks and these chunks being transformed into corresponding embedding vectors by leveraging an embedding model and [0045] for auto-assignment techniques 106 preprocessing the existing hierarchical structure 102 at block 302 that may involve dividing a given hierarchy into smaller sub-hierarchies and as a result, more specific and relevant splits or chunks being generated that are given as input to an embedding model 304 to generate embedding vectors 306. Embedding vectors are generated for the existing sub-structures/chunks/sub-hierarchies.);
generating, at the server, a new vector embedding based on the generated new input metadata [new entity name] (see e.g., [0029] for new entities being input in textual and/or image format that may be first processed to identify names of the new entities using an image identification model such as a machine-learning model, [0031] for processing being performed for new entities that may involve generating new embedding vectors for each new entity by the embedding model, and [0067] for server 1130 being adapted to run one or more services or software applications that enable techniques for auto-assignment of one or more new entities 104. The server generates a new embedding vector based on the new entity name.);
determining, at the server, one or more similarities between the generated new vector embedding and at least one of the generated vector embeddings of the one or more existing database entities (see e.g., [0046] for the new embedding vectors, generated from a query, being compared to the embedding vectors of stored chunks by using similarity metrics such as cosine similarity or Euclidean distance, if the value of a similarity metric for a stored chunk is above a predefined threshold, it being selected and forwarded to a large language model 312 for doing an inference, and based on this comparison, RAG 310 selecting only top k relevant chunks, with embeddings closest to the new embeddings that are associated with the query, to generate a context. The server determines similarities between the new embedding vector and at least one of the embedding vectors of the existing sub-structures/chunks/sub-hierarchies.);
at least one of the existing database entities being determined to be similar to the new database entity when at least one of the corresponding generated vector embeddings is determined to be similar to the corresponding new vector embedding (see e.g., [0046] for the new embedding vectors, generated from a query, being compared to the embedding vectors of stored chunks by using similarity metrics such as cosine similarity or Euclidean distance, if the value of a similarity metric for a stored chunk is above a predefined threshold, it being selected, and selecting only top k relevant chunks, with embeddings closest to the new embeddings that are associated with the query, to generate a context and [0048] for the context from the hierarchical structure simply being the relevant chunks that closely match the new entity name. A hierarchy chunk is determined to be similar to the new entity name when the corresponding chunk embedding vector is determined to be similar to the corresponding new entity name embedding vector.); and
a recommendation to modify at least one of the existing database entities constituting a recommendation to modify at least one of the corresponding generated vector embeddings (see e.g., [0030] for the auto-assignment techniques including dividing the hierarchical structure into smaller sub-structures, or relevant chunks and these chunks being transformed into corresponding embedding vectors by leveraging an embedding model and [0058] for auto-assignment techniques 106 updating (or suggesting) a hierarchical structure 102 based on a query 704, which may include adding or modifying entities within the hierarchical structure 102. It is obvious that a suggestion to modify at least one entity of the hierarchy leads to the modification of the corresponding embedding vector of the hierarchy when the modified hierarchy is submitted to the auto-assignment process because a modified chunk results in a modified embedding vector.).
It would have been obvious to one of ordinary skill in the art prior to the effective filing date of the claimed invention to modify the schedule service of Bansal to generate vector embeddings for the one or more existing database entities; generate, at the server, a new vector embedding based on the generated new input metadata; determine, at the server, one or more similarities between the generated new vector embedding and at least one of the generated vector embeddings of the one or more existing database entities; at least one of the existing database entities being determined to be similar to the new database entity when at least one of the corresponding generated vector embeddings is determined to be similar to the corresponding new vector embedding; and a recommendation to modify at least one of the existing database entities constituting a recommendation to modify at least one of the corresponding generated vector embeddings, as taught by Zhao, for the benefit of providing an efficient comparison and a quick access to the relevant entities (see e.g., Zhao, [0045]).
Bansal in view of Zhao does not specifically disclose denormalizing the retrieved metadata of the one or more existing database entities to extract one or more parameters that include at least one selected from a group consisting of: a data type, a domain set, and a description of a field of the one or more existing database entities; and generating vector embeddings from the denormalized retrieved metadata and the extracted parameters. However, Guerra teaches
denormalizing the retrieved metadata of the one or more existing database entities [target objects] to extract one or more parameters [delimiter-separated values] that include at least one selected from a group consisting of: a data type, a domain set, and a description of a field of the one or more existing database entities (see e.g., col. 4, lines 23-35 for the API 150 dynamically extracting, in runtime, metadata of the target object from a metadata store 154, col. 6, lines 9-16 for each data object 234 encapsulating data and metadata dynamically extracted from one or more database tables and the metadata store and although the database tables and metadata store may contain normalized data (e.g., to reduce data redundancy), the extracted data and metadata being denormalized (e.g., in a delimiter-separated values format) so that they can be more interpretable by the generative AI model, and example metadata stored in the metadata store including text descriptions of the database tables, descriptions of attributes for various fields or columns of those tables, data types for different fields, rules and restrictions on data (e.g., some fields may be labeled non-changeable, etc.), col. 11, line 56 - col. 12, line 16 for example metadata stored in the metadata store including text descriptions of the database tables, descriptions of attributes for various fields or columns of those tables, data types for different fields, rules and restrictions on data (e.g., some fields may be labeled non-changeable, etc.) and extracting the denormalized data including replacing, in runtime, at least some attributes or data values stored in the one or more database tables with respective text descriptions from the metadata store, and col. 12, lines 17-22 for generating, in runtime, a prompt using a selected prompt template and the prompt template including at least one placeholder or parameter which can be replaced with the denormalized data (which can be enriched with descriptive metadata) extracted from the database when assembling the prompt. Metadata of target objects is retrieved from the metadata store. The metadata is denormalized to extract delimiter-separated values, which include descriptions of fields of the target objects.); and
generating vector embeddings from the denormalized retrieved metadata and the extracted parameters (see e.g., col. 7, lines 27-32 for the LLM 300 having an encoder 320 and the encoder 320 processing input text, col. 7, lines 44-56 for the input embedding unit 302 tokenizing a text input into a sequence of tokens, each of which represents a word or part of a word and each token then being mapped to a fixed-length vector known as an input embedding, which provides a continuous representation that captures the meaning and context of the text input, col. 12, lines 17-22 for generating, in runtime, a prompt using a selected prompt template and the prompt template including at least one placeholder or parameter which can be replaced with the denormalized data (which can be enriched with descriptive metadata) extracted from the database when assembling the prompt, and col. 12, lines 39-40 for prompting, in runtime, a generative AI model such as an LLM using the prompt.).
It would have been obvious to one of ordinary skill in the art prior to the effective filing date of the claimed invention to modify the schedule service of Bansal in view of Zhao to denormalize the retrieved metadata of the one or more existing database entities to extract one or more parameters that include at least one selected from a group consisting of: a data type, a domain set, and a description of a field of the one or more existing database entities; and generating vector embeddings from the denormalized retrieved metadata and the extracted parameters, as taught by Guerra, for the benefit of the metadata being more interpretable for the generative AI model (see e.g., Guerra, col. 6, lines 9-16).
Bansal and Zhao in view of Guerra does not specifically disclose wherein the denormalizing comprises combining the retrieved metadata from the existing entities before extracting the parameters. However, Farre Guiu teaches
wherein the denormalizing comprises combining the retrieved metadata from the existing entities [media assets] before extracting the parameters (see e.g., [0016] for flowchart 370 beginning with extracting multiple types of metadata (hereinafter “metadata types”) from a media asset (action 372) and extracting metadata from either of media assets 150 or 152, using metadata extractor 222, [0017] for metadata extractor 222 of metadata extraction and management unit 120/220 including multiple metadata extraction modules corresponding respectively to the metadata types to be extracted from media asset 150/152, [0019] for some or all of the metadata extraction being performed sequentially, and being prioritized based on metadata type and when sequential extraction of metadata types is preferred, selection of which of metadata extraction modules 262 through 266 is used to extract metadata from media asset 150/152 next, as well as determination of extraction parameters for the selected metadata extraction module, being based on an earlier extracted metadata of another type, [0020] for executing metadata extraction and management unit 120/220 to extract and manage metadata for multiple media assets substantially concurrently, [0021] for flowchart 370 continuing with aggregating the multiple metadata types to produce an aggregated metadata describing media asset 150/152 (action 374), and [0023] for transforming the aggregated metadata into one or more database entries 228 identifying media asset 150/152 (action 376). Metadata is retrieved from media assets. The retrieved metadata is combined into aggregated metadata. Then, parameters are extracted in order to retrieve other types of metadata.).
It would have been obvious to one of ordinary skill in the art prior to the effective filing date of the claimed invention to modify the schedule service of Bansal and Zhao in view of Guerra wherein the denormalizing comprises combining the retrieved metadata from the existing entities before extracting the parameters, as taught by Farre Guiu, for the benefit of validating some or all of the metadata included in the second metadata type using metadata included in the fourth metadata type (see e.g., Farre Guiu, [0022]).
As to claim 3, the limitations of parent claim 1 have been discussed above. Bansal does not specifically disclose storing the generated vector embeddings in a storage device that is communicatively coupled to the server. However, Zhao teaches
storing the generated vector embeddings in a storage device [vector database 308] that is communicatively coupled to the server (see e.g., [0045] for more specific and relevant splits or chunks being generated that are given as input to an embedding model 304 to generate embedding vectors 306, which are high-dimensional numerical representations that can be stored, efficiently queried and retrieved from a vector database 308 (e.g., Oracle DB 23ai, Chroma or Milvus database)).
It would have been obvious to one of ordinary skill in the art prior to the effective filing date of the claimed invention to modify the schedule service of Bansal to store the generated vector embeddings in a storage device that is communicatively coupled to the server, as taught by Zhao, for the benefit of providing an efficient comparison and a quick access to the relevant entities (see e.g., Zhao, [0045]).
As to claim 4, the limitations of parent claim 1 have been discussed above. Bansal does not specifically disclose embedding, at the server, the data model attributes into a prompt; and transmitting, at the server, the prompt to a generative artificial intelligence system that is communicatively coupled to the server. However, Zhao teaches
embedding, at the server, the data model attributes [hierarchy relationships, rules, and exceptions] into a prompt (see e.g., [0045] for auto-assignment techniques 106 preprocessing the existing hierarchical structure 102 at block 302 that may involve dividing a given hierarchy into smaller sub-hierarchies, as a result, more specific and relevant splits or chunks being generated that are given as input to an embedding model 304 to generate embedding vectors, and the embedding vectors 306 capturing individual categories, subcategories, semantic relationships between them, the rules to categorize entities, or exceptions allowing to understand the chunks for subsequent processing, [0046] for selecting only top k relevant chunks, with embeddings closest to the new embeddings that are associated with the query, to generate a context and in some instances, RAG 310 employing embedding models like transformers with encoder-decoder architectures (e.g., T5 or GPT) to generate additional text, which may include categories, subcategories, entity attributes, and categorization rules that are highly relevant to the context of a query, and [0047] for human-readable prompt 316 including the query augmented with a context from the existing hierarchical structure 102. The server embeds hierarchy relationships, rules, and exceptions into a prompt.); and
transmitting, at the server, the prompt to a generative artificial intelligence system that is communicatively coupled to the server (see e.g., [0044] for retrieval-augmented generation (RAG) 310 offering a mechanism for augmenting user queries with the context from existing hierarchical structure 102 and an additional context from knowledge base 314 for prompting large language model 312 and/or large multimodal models (LMMs)).
It would have been obvious to one of ordinary skill in the art prior to the effective filing date of the claimed invention to modify the schedule service of Bansal to embed, at the server, the data model attributes into a prompt; and transmitting, at the server, the prompt to a generative artificial intelligence system that is communicatively coupled to the server, as taught by Zhao, for the benefit of producing responses that are not only accurate but also contextually relevant (see e.g., Zhao, [0046]).
As to claim 5, the limitations of parent claims 1 and 4 have been discussed above. Bansal does not specifically disclose wherein the data model attributes embedded in the prompt include instructions for generating the new input metadata. However, Zhao teaches
wherein the data model attributes embedded in the prompt include instructions for generating the new input metadata (see e.g., [0048] for the system prompt instructing the large language model 312 to create the hierarchical level values for the new entities 104 from the top hierarchical level to the bottom (or vice versa), while maintaining adherence of each level to the logic of the existing hierarchical structure 102, the resultant system prompt for this hierarchy extension being of the form: “Generate a complete hierarchy for a ‘Hybrid memory cube (HMC)’ starting from the lowest level to the highest level by following specific rules of categorization of entities in an input hierarchical structure. If a new entity is not semantically similar to any of the existing categories, generate new categories at one or more hierarchical levels and add the new entity under these new categories,” and in some instances, the context from the hierarchical structure may simply be the relevant chunks that closely match the new entity name. The hierarchy relationships, rules, and exceptions embedded in the prompt include instructions for generating the new entity name.).
It would have been obvious to one of ordinary skill in the art prior to the effective filing date of the claimed invention to modify the schedule service of Bansal wherein the data model attributes embedded in the prompt include instructions for generating the new input metadata, as taught by Zhao, for the benefit of producing responses that are not only accurate but also contextually relevant (see e.g., Zhao, [0046]).
As to claim 6, the limitations of parent claim 1 have been discussed above. Bansal does not specifically disclose determining a highest similarity score between the generated new vector embedding and at least one of the generated vector embeddings. However, Zhao teaches
determining a highest similarity score [similarity metric] between the generated new vector embedding and at least one of the generated vector embeddings (see e.g., [0046] for the new embedding vectors, generated from a query, being compared to the embedding vectors of stored chunks by using similarity metrics such as cosine similarity or Euclidean distance, if the value of a similarity metric for a stored chunk is above a predefined threshold, it being selected and forwarded to a large language model 312 for doing an inference, and based on this comparison, RAG 310 selecting only top k relevant chunks, with embeddings closest to the new embeddings that are associated with the query, to generate a context. Selecting the top k relevant chunks includes determining the highest similarity metric between the new embedding vector and the embedding vectors of stored chunks.).
It would have been obvious to one of ordinary skill in the art prior to the effective filing date of the claimed invention to modify the schedule service of Bansal to determine a highest similarity score between the generated new vector embedding and at least one of the generated vector embeddings, as taught by Zhao, for the benefit of providing an efficient comparison and a quick access to the relevant entities (see e.g., Zhao, [0045]).
As to claim 7, the limitations of parent claims 1 and 6 have been discussed above. Bansal does not specifically disclose determining whether the highest similarity score is greater than or equal to a predetermined threshold score for similarity. However, Zhao teaches
determining whether the highest similarity score is greater than or equal to a predetermined threshold score [predefined threshold] for similarity (see e.g., [0046] for the new embedding vectors, generated from a query, being compared to the embedding vectors of stored chunks by using similarity metrics such as cosine similarity or Euclidean distance, if the value of a similarity metric for a stored chunk is above a predefined threshold, it being selected and forwarded to a large language model 312 for doing an inference, and based on this comparison, RAG 310 selecting only top k relevant chunks, with embeddings closest to the new embeddings that are associated with the query, to generate a context. The top similarity metric is determined to be greater than the predefined threshold.).
It would have been obvious to one of ordinary skill in the art prior to the effective filing date of the claimed invention to modify the schedule service of Bansal to determine whether the highest similarity score is greater than or equal to a predetermined threshold score for similarity, as taught by Zhao, for the benefit of providing an efficient comparison and a quick access to the relevant entities (see e.g., Zhao, [0045]).
As to claim 8, the limitations of parent claim 1 have been discussed above. Bansal teaches
generating the recommendation for adding to or modifying the at least one of the existing database entities that is determined to be similar (see e.g., [0075] for new event identifier 272 performing the analysis to determine whether a detected event is new by comparing a set of features or attributes of the newly detected event to features or attributes of previously detected events associated with the user, where the similarity score satisfies a threshold, this implying the newly detected event is sufficiently similar to a previously detected event, and thus is not new, and new event identifier 272 utilizing the enriched information from event content enricher 268 to increase the accuracy of identifying whether an event is new, [0099] for new event integrator 270 determining that an event data object does not correspond to a new event, and updated event determiner 282 determining that an event comprises an update and which event data object(s) should be updated, [0101] for comparing an update entity with existing event entities of the same event type, and then determining whether attributes in the update entity match attributes in the existing event entities of that type, in order to determine the specific existing entity for which the update regards, the attributes or features used for the comparison comprising extracted or supplemental event content, and where the similarity score satisfies a threshold, the one or more entities being determined to be associated with the update entity, and [0102] for event entity reconciliation 284 being generally responsible for reconciling information from an update entity and information from existing event entities for which the update regards, the content in the existing event entity being modified based on the content from the update entity, and the modification comprising changing existing attributes in the existing event entity or adding to the existing event entity additional or new information from the update entity. The updated event determiner’s decision that an existing event entity is associated with an update entity constitutes a recommendation for adding to or modifying the existing event entity.); and
transmitting the generated recommendation to the server (see e.g., [0075] for new event identifier 272 performing the analysis to determine whether a detected event is new by comparing a set of features or attributes of the newly detected event to features or attributes of previously detected events associated with the user, where the similarity score satisfies a threshold, this implying the newly detected event is sufficiently similar to a previously detected event, and thus is not new, and new event identifier 272 utilizing the enriched information from event content enricher 268 to increase the accuracy of identifying whether an event is new, [0099] for new event integrator 270 determining that an event data object does not correspond to a new event, and updated event determiner 282 determining that an event comprises an update and which event data object(s) should be updated, [0101] for comparing an update entity with existing event entities of the same event type, and then determining whether attributes in the update entity match attributes in the existing event entities of that type, in order to determine the specific existing entity for which the update regards, the attributes or features used for the comparison comprising extracted or supplemental event content, and where the similarity score satisfies a threshold, the one or more entities being determined to be associated with the update entity, and [0102] for event entity reconciliation 284 being generally responsible for reconciling information from an update entity and information from existing event entities for which the update regards, the content in the existing event entity being modified based on the content from the update entity, and the modification comprising changing existing attributes in the existing event entity or adding to the existing event entity additional or new information from the update entity. The decision that an existing event entity is associated with an update entity is transmitted to the server, which implements event entity reconciliation 286, so that the update entity and existing event entity can be reconciled.).
Bansal does not specifically disclose generating, at a generative artificial intelligence system communicatively coupled to the server, the recommendation for adding to or modifying the at least one of the generated vector embeddings. However, Zhao teaches
generating, at a generative artificial intelligence system [LLM] communicatively coupled to the server, the recommendation for adding to or modifying the at least one of the generated vector embeddings (see e.g., [0058] for auto-assignment techniques 106 updating (or suggesting) a hierarchical structure 102 based on a query 704, which may include adding or modifying entities within the hierarchical structure 102 and [0059] for with the relevant chunks from the vector database that may contribute to building the context and the additional context from the knowledge base 314, the RAG 310 generating a system prompt 702, the system prompt 702 including instructions for large language model (LLM) 312 to generate or update the hierarchical structure 102 based on the one or more new entities that are mentioned in the query 704, the human-readable prompt 31 including relevant chunks, the context from the hierarchical structure 102, additional context from the knowledge base 314 and the system prompt 702, and subsequently, the human-readable prompt 316 being given as input to large language model 312, that processes different information elements, included in the prompt, to generate an updated or suggested structure 108, which includes the one or more new entities as specified in the query 704. The LLM generates a recommendation for adding to or modifying an entity of the hierarchy, which leads to a modified corresponding embedding vector).
It would have been obvious to one of ordinary skill in the art prior to the effective filing date of the claimed invention to modify the schedule service of Bansal to generate, at a generative artificial intelligence system communicatively coupled to the server, the recommendation for adding to or modifying the at least one of the generated vector embeddings, as taught by Zhao, for the benefit of producing responses that are not only accurate but also contextually relevant (see e.g., Zhao, [0046]).
As to claim 9, the limitations of parent claim 1 have been discussed above. Bansal teaches
generating the recommendation for leaving unchanged the at least one of the existing database entities that is determined to be similar (see e.g., [0099] for new event integrator 270 determining that an event data object does not correspond to a new event, and updated event determiner 282 determining that an event comprises an update and which event data object(s) should be updated and [0102] for event entity reconciliation 284 being generally responsible for reconciling information from an update entity and information from existing event entities for which the update regards and attributes (or data fields for an attribute in an event data object) being compared between the existing entity and the update entity, if either entity has any information that is missing from or conflicting, then event entity reconciliation 284 importing the missing information or modifying the conflicting information. If the existing event entity has no missing information or conflicting information when compared to the update entity, a determination is made that the event entity reconciliation should leave the existing event entity unchanged.).
Bansal does not specifically disclose generating, at a generative artificial intelligence system communicatively coupled to the server, the recommendation for leaving unchanged the at least one of the generated vector embeddings; and transmitting the generated recommendation to the server. However, Zhao teaches
generating, at a generative artificial intelligence system [LLM] communicatively coupled to the server, the recommendation for leaving unchanged the at least one of the generated vector embeddings (see e.g., [0058] for auto-assignment techniques 106 updating (or suggesting) a hierarchical structure 102 based on a query 704, which may include adding or modifying entities within the hierarchical structure 102 and [0059] for with the relevant chunks from the vector database that may contribute to building the context and the additional context from the knowledge base 314, the RAG 310 generating a system prompt 702, the system prompt 702 including instructions for large language model (LLM) 312 to generate or update the hierarchical structure 102 based on the one or more new entities that are mentioned in the query 704, the human-readable prompt 31 including relevant chunks, the context from the hierarchical structure 102, additional context from the knowledge base 314 and the system prompt 702, and subsequently, the human-readable prompt 316 being given as input to large language model 312, that processes different information elements, included in the prompt, to generate an updated or suggested structure 108, which includes the one or more new entities as specified in the query 704. The LLM generates a recommendation for an updated hierarchical structure, which include both modified and unmodified entities. Unmodified entities lead to unmodified embedding vectors.); and
transmitting the generated recommendation to the server (see e.g., [0034] for various evaluations being performed on the suggested hierarchical structure, involving quality control and error measurements and [0067] for server 1130 being adapted to run one or more services or software applications that enable techniques for auto-assignment of one or more new entities 104. The recommended hierarchy, including the modified and unmodified entities, is transmitted to the server for evaluation.).
It would have been obvious to one of ordinary skill in the art prior to the effective filing date of the claimed invention to modify the schedule service of Bansal to generate, at a generative artificial intelligence system communicatively coupled to the server, the recommendation for leaving unchanged the at least one of the generated vector embeddings; and transmitting the generated recommendation to the server, as taught by Zhao, for the benefit of producing responses that are not only accurate but also contextually relevant (see e.g., Zhao, [0046]).
As to claim 10, Bansal recites a system comprising:
a server communicatively coupled to a database system (see e.g., [0028] for an event data object (sometimes referred to as an event “entity”) corresponding to a first event being stored in association with another entity by utilizing a relational database, [0034] for server 106 being provided via multiple devices arranged in a distributed environment that collectively provide the functionality described herein, and [0035] for server 106 comprising server-side software designed to work in conjunction with client-side software on user devices 102a and 102b through 102n so as to implement any combination of the features and functionalities discussed in the present disclosure), the server comprising at least one processor and the server configured to:
retrieve metadata [event features/attributes] of one or more existing database entities [event data objects] of the database system (see e.g., [0034] for server 106 being provided via multiple devices arranged in a distributed environment that collectively provide the functionality described herein, [0035] for server 106 comprising server-side software designed to work in conjunction with client-side software on user devices 102a and 102b through 102n so as to implement any combination of the features and functionalities discussed in the present disclosure, [0061] for event content extractor 266 then extracting event content from information about the user-related activity and event content generally referring to any information associated with an event, including but not limited to information about a service provider, location, name of event, date, event attendees, confirmation or reservation information, tracking or itinerary numbers, information specific to event types such as flight numbers, check-in/pick-up and check-out/drop-off times, agendas, or other details that are specific to a particular event, [0063] for event content extractor 266 also generating an indication or record of the event, which may comprise an event data object, and which may be stored in event data objects (entities) 242, a user profile 240, or otherwise stored and associated with the user, [0064] for event content enricher 268, in general, enriching event data objects, or the event content, with supplemental information that is relevant to that particular event and supplemental information including contextual information for the event that is not expressly included in a particular instance of user-related activity, for example and without limitation, supplemental information including: enriched location-related content (e.g., determining geo-location from extracted content, such as determining city, region, latitude-longitude from airport codes for travel); contact information and other information about service providers associated with an event (e.g., a hotel's phone number, email, or website); or nearly any other information related to the event, and extracted content corresponding to an event, or aspects of the user-related activity information associated with the event being utilized by event content enricher 268 to generate supplemental information, [0065] for new event integrator 270 utilizing the supplemental information for determining whether an event is new or related to other events, [0074] for new event integrator 270 comprising a new event identifier 272, new event identifier 272 receiving information about an event, such as an indication of the event or an event data object for the event, and analyzing the information to determine whether the corresponding event is new, such as by comparing the information to other information about the other, previously detected events associated with the user, and information about the previously detected events being determined or received from event data objects corresponding to those events, which may be stored in a user profile 240 associated with the user, such as in event data objects (entities) 242, and [0075] for new event identifier 272 performing the analysis to determine whether a detected event is new by comparing a set of features or attributes of the newly detected event to features or attributes of previously detected events associated with the user. Server 106, which may implement the new event integrator 270, retrieves event features/attributes of an existing event data object. The event features/attributes include context information corresponding to the event data object.);
receive data model attributes for a new database entity and generate new input metadata based on the data model attributes (see e.g., [0025] for the event content being identified utilizing data models, [0116] for the term “data object” or “entity” being used broadly herein and comprising a record or set of records, data organized according to a data model (which may include a set of one or more data objects), a data structure, or template (which may be pre-defined and/or may be based on the particular category of event), a class object, a combination of these, or similar data structure and for instance, in some embodiments, the data structure comprising a pre-defined structure for each event type or being defined from a data structure template or data model, [0063] for event content extractor 266 generating an event data object (or an event entity) corresponding to a particular event that comprises the extracted event content from user-related activity information indicating or associated with the event and in some embodiments, event content extractor 266 providing the event data object, or extracted event content, to event content enricher 268, [0064] for event content enricher 268, in general, enriching event data objects, or the event content, with supplemental information that is relevant to that particular event, and [0074] for new event identifier 272 receiving information about an event, such as an indication of the event or an event data object for the event, and analyzing the information to determine whether the corresponding event is new, such as by comparing the information to other information about the other, previously detected events associated with the user. The server, which implements event detector 260 and new event integrator 270, receives data model attributes for the purpose of providing an organized structure for new event data objects. New event content extraction data and event content enrichment data are generated based on the model attributes. These new event features/attributes are received by the new event identifier for comparison purposes.); and
receiving, at the sever, a recommendation of whether at least one of the existing database entities that is determined to be similar to the new database entity is to be added to or modified based on the generated new input metadata (see e.g., [0075] for new event identifier 272 performing the analysis to determine whether a detected event is new by comparing a set of features or attributes of the newly detected event to features or attributes of previously detected events associated with the user, where the similarity score satisfies a threshold, this implying the newly detected event is sufficiently similar to a previously detected event, and thus is not new, and new event identifier 272 utilizing the enriched information from event content enricher 268 to increase the accuracy of identifying whether an event is new, [0099] for new event integrator 270 determining that an event data object does not correspond to a new event, and updated event determiner 282 determining that an event comprises an update and which event data object(s) should be updated, [0101] comparing an update entity with existing event entities of the same event type, and then determining whether attributes in the update entity match attributes in the existing event entities of that type, in order to determine the specific existing entity for which the update regards, the attributes or features used for the comparison comprising extracted or supplemental event content, and where the similarity score satisfies a threshold, the one or more entities being determined to be associated with the update entity, and [0102] for event entity reconciliation 284 being generally responsible for reconciling information from an update entity and information from existing event entities for which the update regards, the content in the existing event entity being modified based on the content from the update entity, and the modification comprising changing existing attributes in the existing event entity or adding to the existing event entity additional or new information from the update entity. The server, which implements event entity reconciliation 286, receives a recommendation of whether at least one of the existing event data objects that is determined to be similar to the new event data object is to be added to or modified based on the newly generated event content extraction data and event content enrichment data.).
Bansal does not specifically disclose generate vector embeddings for the one or more existing database entities; generate a new vector embedding based on the generated new input metadata; determine one or more similarities between the generated new vector embedding and at least one of the generated vector embeddings of the one or more existing database entities; at least one of the existing database entities being determined to be similar to the new database entity when at least one of the corresponding generated vector embeddings is determined to be similar to the corresponding new vector embedding; and a recommendation to modify at least one of the existing database entities constituting a recommendation to modify at least one of the corresponding generated vector embeddings. However, Zhao teaches
generate vector embeddings for the one or more existing database entities [sub-structures/chunks/sub-hierarchies] (see e.g., [0030] for the auto-assignment techniques including dividing the hierarchical structure into smaller sub-structures, or relevant chunks and these chunks being transformed into corresponding embedding vectors by leveraging an embedding model and [0045] for auto-assignment techniques 106 preprocessing the existing hierarchical structure 102 at block 302 that may involve dividing a given hierarchy into smaller sub-hierarchies and as a result, more specific and relevant splits or chunks being generated that are given as input to an embedding model 304 to generate embedding vectors 306. Embedding vectors are generated for the existing sub-structures/chunks/sub-hierarchies.);
generate a new vector embedding based on the generated new input metadata [new entity name] (see e.g., [0029] for new entities being input in textual and/or image format that may be first processed to identify names of the new entities using an image identification model such as a machine-learning model, [0031] for processing being performed for new entities that may involve generating new embedding vectors for each new entity by the embedding model, and [0067] for server 1130 being adapted to run one or more services or software applications that enable techniques for auto-assignment of one or more new entities 104. The server generates a new embedding vector based on the new entity name.);
determine one or more similarities between the generated new vector embedding and at least one of the generated vector embeddings of the one or more existing database entities (see e.g., [0046] for the new embedding vectors, generated from a query, being compared to the embedding vectors of stored chunks by using similarity metrics such as cosine similarity or Euclidean distance, if the value of a similarity metric for a stored chunk is above a predefined threshold, it being selected and forwarded to a large language model 312 for doing an inference, and based on this comparison, RAG 310 selecting only top k relevant chunks, with embeddings closest to the new embeddings that are associated with the query, to generate a context. The server determines similarities between the new embedding vector and at least one of the embedding vectors of the existing sub-structures/chunks/sub-hierarchies.);
at least one of the existing database entities being determined to be similar to the new database entity when at least one of the corresponding generated vector embeddings is determined to be similar to the corresponding new vector embedding (see e.g., [0046] for the new embedding vectors, generated from a query, being compared to the embedding vectors of stored chunks by using similarity metrics such as cosine similarity or Euclidean distance, if the value of a similarity metric for a stored chunk is above a predefined threshold, it being selected, and selecting only top k relevant chunks, with embeddings closest to the new embeddings that are associated with the query, to generate a context and [0048] for the context from the hierarchical structure simply being the relevant chunks that closely match the new entity name. A hierarchy chunk is determined to be similar to the new entity name when the corresponding chunk embedding vector is determined to be similar to the corresponding new entity name embedding vector.); and
a recommendation to modify at least one of the existing database entities constituting a recommendation to modify at least one of the corresponding generated vector embeddings (see e.g., [0030] for the auto-assignment techniques including dividing the hierarchical structure into smaller sub-structures, or relevant chunks and these chunks being transformed into corresponding embedding vectors by leveraging an embedding model and [0058] for auto-assignment techniques 106 updating (or suggesting) a hierarchical structure 102 based on a query 704, which may include adding or modifying entities within the hierarchical structure 102. It is obvious that a suggestion to modify at least one entity of the hierarchy leads to the modification of the corresponding embedding vector of the hierarchy when the modified hierarchy is submitted to the auto-assignment process because a modified chunk results in a modified embedding vector.).
It would have been obvious to one of ordinary skill in the art prior to the effective filing date of the claimed invention to modify the schedule service of Bansal to generate vector embeddings for the one or more existing database entities; generate a new vector embedding based on the generated new input metadata; determine one or more similarities between the generated new vector embedding and at least one of the generated vector embeddings of the one or more existing database entities;
at least one of the existing database entities being determined to be similar to the new database entity when at least one of the corresponding generated vector embeddings is determined to be similar to the corresponding new vector embedding; and a recommendation to modify at least one of the existing database entities constituting a recommendation to modify at least one of the corresponding generated vector embeddings, as taught by Zhao, for the benefit of providing an efficient comparison and a quick access to the relevant entities (see e.g., Zhao, [0045]).
Bansal in view of Zhao does not specifically disclose denormalize the retrieved metadata of the one or more existing database entities to extract one or more parameters that include at least one selected from a group consisting of: a data type, a domain set, and a description of a field of the one or more existing database entities; and generate vector embeddings from the denormalized retrieved metadata and the extracted parameters. However, Guerra teaches
denormalize the retrieved metadata of the one or more existing database entities [target objects] to extract one or more parameters [delimiter-separated values] that include at least one selected from a group consisting of: a data type, a domain set, and a description of a field of the one or more existing database entities (see e.g., col. 4, lines 23-35 for the API 150 dynamically extracting, in runtime, metadata of the target object from a metadata store 154, col. 6, lines 9-16 for each data object 234 encapsulating data and metadata dynamically extracted from one or more database tables and the metadata store and although the database tables and metadata store may contain normalized data (e.g., to reduce data redundancy), the extracted data and metadata being denormalized (e.g., in a delimiter-separated values format) so that they can be more interpretable by the generative AI model, and example metadata stored in the metadata store including text descriptions of the database tables, descriptions of attributes for various fields or columns of those tables, data types for different fields, rules and restrictions on data (e.g., some fields may be labeled non-changeable, etc.), col. 11, line 56 - col. 12, line 16 for example metadata stored in the metadata store including text descriptions of the database tables, descriptions of attributes for various fields or columns of those tables, data types for different fields, rules and restrictions on data (e.g., some fields may be labeled non-changeable, etc.) and extracting the denormalized data including replacing, in runtime, at least some attributes or data values stored in the one or more database tables with respective text descriptions from the metadata store, and col. 12, lines 17-22 for generating, in runtime, a prompt using a selected prompt template and the prompt template including at least one placeholder or parameter which can be replaced with the denormalized data (which can be enriched with descriptive metadata) extracted from the database when assembling the prompt. Metadata of target objects is retrieved from the metadata store. The metadata is denormalized to extract delimiter-separated values, which include descriptions of fields of the target objects.); and
generate vector embeddings from the denormalized retrieved metadata and the extracted parameters (see e.g., col. 7, lines 27-32 for the LLM 300 having an encoder 320 and the encoder 320 processing input text, col. 7, lines 44-56 for the input embedding unit 302 tokenizing a text input into a sequence of tokens, each of which represents a word or part of a word and each token then being mapped to a fixed-length vector known as an input embedding, which provides a continuous representation that captures the meaning and context of the text input, col. 12, lines 17-22 for generating, in runtime, a prompt using a selected prompt template and the prompt template including at least one placeholder or parameter which can be replaced with the denormalized data (which can be enriched with descriptive metadata) extracted from the database when assembling the prompt, and col. 12, lines 39-40 for prompting, in runtime, a generative AI model such as an LLM using the prompt.).
It would have been obvious to one of ordinary skill in the art prior to the effective filing date of the claimed invention to modify the schedule service of Bansal in view of Zhao to denormalize the retrieved metadata of the one or more existing database entities to extract one or more parameters that include at least one selected from a group consisting of: a data type, a domain set, and a description of a field of the one or more existing database entities; and generate vector embeddings from the denormalized retrieved metadata and the extracted parameters, as taught by Guerra, for the benefit of the metadata being more interpretable for the generative AI model (see e.g., Guerra, col. 6, lines 9-16).
Bansal and Zhao in view of Guerra does not specifically disclose wherein the denormalizing comprises combining the retrieved metadata from the existing entities before extracting the parameters. However, Farre Guiu teaches
wherein the denormalizing comprises combining the retrieved metadata from the existing entities [media assets] before extracting the parameters (see e.g., [0016] for flowchart 370 beginning with extracting multiple types of metadata (hereinafter “metadata types”) from a media asset (action 372) and extracting metadata from either of media assets 150 or 152, using metadata extractor 222, [0017] for metadata extractor 222 of metadata extraction and management unit 120/220 including multiple metadata extraction modules corresponding respectively to the metadata types to be extracted from media asset 150/152, [0019] for some or all of the metadata extraction being performed sequentially, and being prioritized based on metadata type and when sequential extraction of metadata types is preferred, selection of which of metadata extraction modules 262 through 266 is used to extract metadata from media asset 150/152 next, as well as determination of extraction parameters for the selected metadata extraction module, being based on an earlier extracted metadata of another type, [0020] for executing metadata extraction and management unit 120/220 to extract and manage metadata for multiple media assets substantially concurrently, [0021] for flowchart 370 continuing with aggregating the multiple metadata types to produce an aggregated metadata describing media asset 150/152 (action 374), and [0023] for transforming the aggregated metadata into one or more database entries 228 identifying media asset 150/152 (action 376). Metadata is retrieved from media assets. The retrieved metadata is combined into aggregated metadata. Then, parameters are extracted in order to retrieve other types of metadata.).
It would have been obvious to one of ordinary skill in the art prior to the effective filing date of the claimed invention to modify the schedule service of Bansal and Zhao in view of Guerra wherein the denormalizing comprises combining the retrieved metadata from the existing entities before extracting the parameters, as taught by Farre Guiu, for the benefit of validating some or all of the metadata included in the second metadata type using metadata included in the fourth metadata type (see e.g., Farre Guiu, [0022]).
As to claim 12, the limitations of parent claim 10 have been discussed above. Bansal does not specifically disclose a storage device communicatively coupled to the server, wherein the server is configured to store the generated vector embeddings in the storage device. However, Zhao teaches
a storage device [vector database 308] communicatively coupled to the server,
wherein the server is configured to store the generated vector embeddings in the storage device (see e.g., [0045] for more specific and relevant splits or chunks being generated that are given as input to an embedding model 304 to generate embedding vectors 306, which are high-dimensional numerical representations that can be stored, efficiently queried and retrieved from a vector database 308 (e.g., Oracle DB 23ai, Chroma or Milvus database)).
It would have been obvious to one of ordinary skill in the art prior to the effective filing date of the claimed invention to modify the schedule service of Bansal to include a storage device communicatively coupled to the server, wherein the server is configured to store the generated vector embeddings in the storage device, as taught by Zhao, for the benefit of providing an efficient comparison and a quick access to the relevant entities (see e.g., Zhao, [0045]).
As to claim 13, the limitations of parent claim 10 have been discussed above. Bansal does not specifically disclose wherein the server is configured to embed the data model attributes into a prompt, and transmit the prompt to a generative artificial intelligence system that is communicatively coupled to the server. However, Zhao teaches
wherein the server is configured to embed the data model attributes [hierarchy relationships, rules, and exceptions] into a prompt (see e.g., [0045] for auto-assignment techniques 106 preprocessing the existing hierarchical structure 102 at block 302 that may involve dividing a given hierarchy into smaller sub-hierarchies, as a result, more specific and relevant splits or chunks being generated that are given as input to an embedding model 304 to generate embedding vectors, and the embedding vectors 306 capturing individual categories, subcategories, semantic relationships between them, the rules to categorize entities, or exceptions allowing to understand the chunks for subsequent processing, [0046] for selecting only top k relevant chunks, with embeddings closest to the new embeddings that are associated with the query, to generate a context and in some instances, RAG 310 employing embedding models like transformers with encoder-decoder architectures (e.g., T5 or GPT) to generate additional text, which may include categories, subcategories, entity attributes, and categorization rules that are highly relevant to the context of a query, and [0047] for human-readable prompt 316 including the query augmented with a context from the existing hierarchical structure 102. The server embeds hierarchy relationships, rules, and exceptions into a prompt.), and transmit at the prompt to a generative artificial intelligence system that is communicatively coupled to the server (see e.g., [0044] for retrieval-augmented generation (RAG) 310 offering a mechanism for augmenting user queries with the context from existing hierarchical structure 102 and an additional context from knowledge base 314 for prompting large language model 312 and/or large multimodal models (LMMs)).
It would have been obvious to one of ordinary skill in the art prior to the effective filing date of the claimed invention to modify the schedule service of Bansal wherein the server is configured to embed the data model attributes into a prompt, and transmit the prompt to a generative artificial intelligence system that is communicatively coupled to the server, as taught by Zhao, for the benefit of producing responses that are not only accurate but also contextually relevant (see e.g., Zhao, [0046]).
As to claim 14, the limitations of parent claims 10 and 13 have been discussed above. Bansal does not specifically disclose wherein the data model attributes embedded in the prompt include instructions for generating the new input metadata. However, Zhao teaches
wherein the data model attributes embedded in the prompt include instructions for generating the new input metadata (see e.g., [0048] for the system prompt instructing the large language model 312 to create the hierarchical level values for the new entities 104 from the top hierarchical level to the bottom (or vice versa), while maintaining adherence of each level to the logic of the existing hierarchical structure 102, the resultant system prompt for this hierarchy extension being of the form: “Generate a complete hierarchy for a ‘Hybrid memory cube (HMC)’ starting from the lowest level to the highest level by following specific rules of categorization of entities in an input hierarchical structure. If a new entity is not semantically similar to any of the existing categories, generate new categories at one or more hierarchical levels and add the new entity under these new categories,” and in some instances, the context from the hierarchical structure may simply be the relevant chunks that closely match the new entity name. The hierarchy relationships, rules, and exceptions embedded in the prompt include instructions for generating the new entity name.).
It would have been obvious to one of ordinary skill in the art prior to the effective filing date of the claimed invention to modify the schedule service of Bansal wherein the data model attributes embedded in the prompt include instructions for generating the new input metadata, as taught by Zhao, for the benefit of producing responses that are not only accurate but also contextually relevant (see e.g., Zhao, [0046]).
As to claim 15, the limitations of parent claim 10 have been discussed above. Bansal does not specifically disclose wherein the server is configured to determine the one or more similarities by determining a highest similarity score between the generated new vector embedding and at least one of the generated vector embeddings. However, Zhao teaches
wherein the server is configured to determine the one or more similarities by determining a highest similarity score [similarity metric] between the generated new vector embedding and at least one of the generated vector embeddings (see e.g., [0046] for the new embedding vectors, generated from a query, being compared to the embedding vectors of stored chunks by using similarity metrics such as cosine similarity or Euclidean distance, if the value of a similarity metric for a stored chunk is above a predefined threshold, it being selected and forwarded to a large language model 312 for doing an inference, and based on this comparison, RAG 310 selecting only top k relevant chunks, with embeddings closest to the new embeddings that are associated with the query, to generate a context. Selecting the top k relevant chunks includes determining the highest similarity metric between the new embedding vector and the embedding vectors of stored chunks.).
It would have been obvious to one of ordinary skill in the art prior to the effective filing date of the claimed invention to modify the schedule service of Bansal to wherein the server is configured to determine the one or more similarities by determining a highest similarity score between the generated new vector embedding and at least one of the generated vector embeddings, as taught by Zhao, for the benefit of providing an efficient comparison and a quick access to the relevant entities (see e.g., Zhao, [0045]).
As to claim 16, the limitations of parent claims 10 and 15 have been discussed above. Bansal does not specifically disclose wherein the server is configured to determine whether the highest similarity score is greater than or equal to a predetermined threshold score for similarity. However, Zhao teaches
wherein the server is configured to determine whether the highest similarity score is greater than or equal to a predetermined threshold score [predefined threshold] for similarity (see e.g., [0046] for the new embedding vectors, generated from a query, being compared to the embedding vectors of stored chunks by using similarity metrics such as cosine similarity or Euclidean distance, if the value of a similarity metric for a stored chunk is above a predefined threshold, it being selected and forwarded to a large language model 312 for doing an inference, and based on this comparison, RAG 310 selecting only top k relevant chunks, with embeddings closest to the new embeddings that are associated with the query, to generate a context. The top similarity metric is determined to be greater than the predefined threshold.).
It would have been obvious to one of ordinary skill in the art prior to the effective filing date of the claimed invention to modify the schedule service of Bansal wherein the server is configured to determine whether the highest similarity score is greater than or equal to a predetermined threshold score for similarity, as taught by Zhao, for the benefit of providing an efficient comparison and a quick access to the relevant entities (see e.g., Zhao, [0045]).
As to claim 17, the limitations of parent claim 10 have been discussed above. Bansal teaches
generating the recommendation for adding to or modifying the at least one of the existing database entities that is determined to be similar (see e.g., [0075] for new event identifier 272 performing the analysis to determine whether a detected event is new by comparing a set of features or attributes of the newly detected event to features or attributes of previously detected events associated with the user, where the similarity score satisfies a threshold, this implying the newly detected event is sufficiently similar to a previously detected event, and thus is not new, and new event identifier 272 utilizing the enriched information from event content enricher 268 to increase the accuracy of identifying whether an event is new, [0099] for new event integrator 270 determining that an event data object does not correspond to a new event, and updated event determiner 282 determining that an event comprises an update and which event data object(s) should be updated, [0101] for comparing an update entity with existing event entities of the same event type, and then determining whether attributes in the update entity match attributes in the existing event entities of that type, in order to determine the specific existing entity for which the update regards, the attributes or features used for the comparison comprising extracted or supplemental event content, and where the similarity score satisfies a threshold, the one or more entities being determined to be associated with the update entity, and [0102] for event entity reconciliation 284 being generally responsible for reconciling information from an update entity and information from existing event entities for which the update regards, the content in the existing event entity being modified based on the content from the update entity, and the modification comprising changing existing attributes in the existing event entity or adding to the existing event entity additional or new information from the update entity. The updated event determiner’s decision that an existing event entity is associated with an update entity constitutes a recommendation for adding to or modifying the existing event entity.) and transmitting the generated recommendation to the server (see e.g., [0075] for new event identifier 272 performing the analysis to determine whether a detected event is new by comparing a set of features or attributes of the newly detected event to features or attributes of previously detected events associated with the user, where the similarity score satisfies a threshold, this implying the newly detected event is sufficiently similar to a previously detected event, and thus is not new, and new event identifier 272 utilizing the enriched information from event content enricher 268 to increase the accuracy of identifying whether an event is new, [0099] for new event integrator 270 determining that an event data object does not correspond to a new event, and updated event determiner 282 determining that an event comprises an update and which event data object(s) should be updated, [0101] for comparing an update entity with existing event entities of the same event type, and then determining whether attributes in the update entity match attributes in the existing event entities of that type, in order to determine the specific existing entity for which the update regards, the attributes or features used for the comparison comprising extracted or supplemental event content, and where the similarity score satisfies a threshold, the one or more entities being determined to be associated with the update entity, and [0102] for event entity reconciliation 284 being generally responsible for reconciling information from an update entity and information from existing event entities for which the update regards, the content in the existing event entity being modified based on the content from the update entity, and the modification comprising changing existing attributes in the existing event entity or adding to the existing event entity additional or new information from the update entity. The decision that an existing event entity is associated with an update entity is transmitted to the server, which implements event entity reconciliation 286, so that the update entity and existing event entity can be reconciled.).
Bansal does not specifically disclose a generative artificial intelligence system communicatively coupled to the server, wherein the generative artificial intelligence system is configured to generate the recommendation for adding to or modifying the at least one of the generated vector embeddings. However, Zhao teaches
a generative artificial intelligence system [LLM] communicatively coupled to the server,
wherein the generative artificial intelligence system is configured to generate the recommendation for adding to or modifying the at least one of the generated vector embeddings (see e.g., [0058] for auto-assignment techniques 106 updating (or suggesting) a hierarchical structure 102 based on a query 704, which may include adding or modifying entities within the hierarchical structure 102 and [0059] for with the relevant chunks from the vector database that may contribute to building the context and the additional context from the knowledge base 314, the RAG 310 generating a system prompt 702, the system prompt 702 including instructions for large language model (LLM) 312 to generate or update the hierarchical structure 102 based on the one or more new entities that are mentioned in the query 704, the human-readable prompt 31 including relevant chunks, the context from the hierarchical structure 102, additional context from the knowledge base 314 and the system prompt 702, and subsequently, the human-readable prompt 316 being given as input to large language model 312, that processes different information elements, included in the prompt, to generate an updated or suggested structure 108, which includes the one or more new entities as specified in the query 704. The LLM generates a recommendation for adding to or modifying an entity of the hierarchy, which leads to a modified corresponding embedding vector).
It would have been obvious to one of ordinary skill in the art prior to the effective filing date of the claimed invention to modify the schedule service of Bansal to include a generative artificial intelligence system communicatively coupled to the server, wherein the generative artificial intelligence system is configured to generate the recommendation for adding to or modifying the at least one of the generated vector embeddings, as taught by Zhao, for the benefit of producing responses that are not only accurate but also contextually relevant (see e.g., Zhao, [0046]).
As to claim 18, the limitations of parent claim 10 have been discussed above. Bansal teaches
generating the recommendation for leaving unchanged the at least one of the existing database entities that is determined to be similar (see e.g., [0099] for new event integrator 270 determining that an event data object does not correspond to a new event, and updated event determiner 282 determining that an event comprises an update and which event data object(s) should be updated and [0102] for event entity reconciliation 284 being generally responsible for reconciling information from an update entity and information from existing event entities for which the update regards and attributes (or data fields for an attribute in an event data object) being compared between the existing entity and the update entity, if either entity has any information that is missing from or conflicting, then event entity reconciliation 284 importing the missing information or modifying the conflicting information. If the existing event entity has no missing information or conflicting information when compared to the update entity, a determination is made that the event entity reconciliation should leave the existing event entity unchanged.).
Bansal does not specifically disclose a generative artificial intelligence system communicatively coupled to the server, wherein the generative artificial intelligence system is configured to generate the recommendation for leaving unchanged the at least one of the generated vector embeddings and transmit the generated recommendation to the server. However, Zhao teaches
a generative artificial intelligence system [LLM] communicatively coupled to the server,
wherein the generative artificial intelligence system is configured to generate the recommendation for leaving unchanged the at least one of the generated vector embeddings (see e.g., [0058] for auto-assignment techniques 106 updating (or suggesting) a hierarchical structure 102 based on a query 704, which may include adding or modifying entities within the hierarchical structure 102 and [0059] for with the relevant chunks from the vector database that may contribute to building the context and the additional context from the knowledge base 314, the RAG 310 generating a system prompt 702, the system prompt 702 including instructions for large language model (LLM) 312 to generate or update the hierarchical structure 102 based on the one or more new entities that are mentioned in the query 704, the human-readable prompt 31 including relevant chunks, the context from the hierarchical structure 102, additional context from the knowledge base 314 and the system prompt 702, and subsequently, the human-readable prompt 316 being given as input to large language model 312, that processes different information elements, included in the prompt, to generate an updated or suggested structure 108, which includes the one or more new entities as specified in the query 704. The LLM generates a recommendation for an updated hierarchical structure, which include both modified and unmodified entities. Unmodified entities lead to unmodified embedding vectors.) and transmit the generated recommendation to the server (see e.g., [0034] for various evaluations being performed on the suggested hierarchical structure, involving quality control and error measurements and [0067] for server 1130 being adapted to run one or more services or software applications that enable techniques for auto-assignment of one or more new entities 104. The recommended hierarchy, including the modified and unmodified entities, is transmitted to the server for evaluation.).
It would have been obvious to one of ordinary skill in the art prior to the effective filing date of the claimed invention to modify the schedule service of Bansal to include a generative artificial intelligence system communicatively coupled to the server, wherein the generative artificial intelligence system is configured to generate the recommendation for leaving unchanged the at least one of the generated vector embeddings and transmit the generated recommendation to the server, as taught by Zhao, for the benefit of producing responses that are not only accurate but also contextually relevant (see e.g., Zhao, [0046]).
Response to Amendment
The 35 U.S.C. 101 rejection of claims 10-18 has been withdrawn in light of Applicant’s amendment to independent claim 10.
Response to Arguments
Applicant’s arguments, see pages 9-11 of Applicant’s Response, filed July 2, 2026, with respect to the rejections of claims 1, 3-10, and 12-18 under 35 U.S.C 103(a) have been fully considered and are persuasive. Therefore, the rejection has been withdrawn. However, upon further consideration, a new ground(s) of rejection is made in view of Farre Guiu et al. (US Publication No. 2017/0109419).
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Zhang (US Publication No. 2021/0383260) for the “denorm charges 618 comprises a set of transactions and associated metadata for such transactions, which are aggregated from multiple tables and then denormalized for storage into a single table” (see [0071]).
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to DARA J GLASSER whose telephone number is (571)270-3666. The examiner can normally be reached Monday-Thursday, 10:00am-2:00pm.
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, Apu Mofiz can be reached at (571)272-4080. 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.
08-20-2026
/DARA J GLASSER/Examiner, Art Unit 2161
/APU M MOFIZ/Supervisory Patent Examiner, Art Unit 2161
Prosecution Insights