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 .
Acknowledgement
Acknowledgement is made of applicant’s amendment made on 12/05/2025. Applicant’s submission filed has been entered and made of record.
Status of the Claims
Claims 1-11, 13, and 15-22 are pending.
Response to Applicant’s Arguments
In response to “Amended claim 1 now recites that the modified user input includes an instruction to only consider the injected triples, or their representation, as a sole factual basis for generating the response to the user input. This amendment introduces a system generated instruction that restricts the factual basis available to the large language model. In other words, the model is expressly directed to treat the injected triples as the exclusive set of facts to be used during the generation of the response” and “Because the amendment in claim 1 addresses the internal attribution and factual basis of the model's response, and because Bhatia never addresses these issues, amended claim 1 is not anticipated”.
In view of such amendment to claims 1 and 18, anticipation rejections under Bhatia are withdrawn. Upon further search and consideration, please see details of a new combination of references set forth below.
In response to “Amended claim 18 introduces a different type of instruction. Claim 18 now recites an instruction directing the large language model to treat the injected triples, or their representation, as system provided new information introduced by the computing system. This claim element affects how the model interprets the provenance of the injected information. In the amended claim, the model is not only given additional content but is instructed to treat that content as new information that originates from the system rather than from the user. Bhatia does not disclose any instruction about the origin, status, or attribution of the injected triples”.
In view of such amendment to claim 20, anticipation rejection under Bhatia has been withdrawn. Upon further search and consideration, please see details of a new combination of references set forth below.
Claim Rejections - 35 USC § 103
The following is a quotation of the appropriate paragraphs of 35 U.S.C. 103 that form the basis for the rejections under this section made in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claims 1-11, 13, 16-20 and 22 are rejected under 35 USC 103(a) as being unpatentable over Bhatia et al. (US 11997056 B2) in view of Bacarella et al. (US 20019/0213258 A1).
Regarding Claims 1, 18, 20, and 22, Bhatia discloses a computing system (Fig. 6, Col 3, Rows 20-22, environment 100 of Fig. 1 operates on one or more computing devices such as computing device 600 per Col 9, Rows 55-58) comprising:
at least one memory (Col 3, Rows 41-43 and Col 10, Rows 33-39, memory / computer readable media);
one or more hardware processing units coupled to the at least one memory (Col 3, Rows 41-43 and Col 10, Rows 15-16, processors 614 have memory per Col 10, Row 25); and
one or more computer readable storage media storing computer-executable instructions that, when executed, cause the computing system (Col 3, Rows 41-43, functions carried out by a processor executing instructions stored in memory) to perform operations comprising:
receiving, from a user, user input (Col 3, Row 63 – Col 4, Row 15, receive natural language question from a user through audio and textual inputs) comprising a plurality of tokens (Col 4, Rows 54-56, provide natural language input 101 to locate and classify named entity comprising one or more words or tokens);
analyzing at least a portion of the plurality of tokens (Col 4, Rows 54-67, perform information extraction on natural language input 101 to locate and classify named entities mentioned in unstructured text into pre-defined entity classes);
based on the analyzing, determining one or more entities of a semantic framework represented in the at least a portion of the plurality of tokens (Col 5, Rows 6-16, identify one or more entities within the natural language input 101 to search and retrieve information from knowledge base 120 for each entity; per Col 5, Rows 26—27, the knowledge base 120 being a knowledge graph);
determining one or more triples of the semantic framework for at least a portion of the one or more entities or for associated entities (Col 5, Rows 40-41, retrieve triples related to the identified entities);
adding at least a portion of the one or more triples, or a representation thereof, to the user input to provide modified user input (Col 6, Rows 1-5, concatenate triple head entity, the relationship, and the tail to form a natural language phrase; Col 6, Rows 15-29, select the top-k triples that constitute contextual knowledge to be fed as input along with input text X to a NLP model 112);
submitting the modified user input to a large language model (Col 6, Rows 15-29, select the top-k triples that constitute contextual knowledge to be fed as input along with input text X to the NLP model 112; per Col 4, Rows 32-33, the NLP model 112 being a BERT, RoBERTa, T5, and GPT);
processing the modified user input using the large language model to provide a response (Col 2, Rows 61-64, NLP model processes both the natural language input and the triples to generate a response; e.g., Col 4, Rows 3-5, search applications receive a natural language query for input and provide one or more results that are responsive to the query); and
returning the response in response to the receiving the user input (Col 4, Rows 3-5, search applications receive a natural language query for input and provide one or more results that are responsive to the query; e.g., Col 4, Rows 9-12).
Bhatia does not disclose adding to the user input to provide modified user input further includes an instruction to only consider the at least a portion of the one or more triples, or a representation thereof, as a sole factual basis for generating a response to the user input and an instruction directing a large language model to treat the added triples, or the representation thereof, as system provided new information introduced by the computing system.
Bacarella discloses a machine learning model system using machine learning models (MLMs) to process natural language / user inputs (Abstract and ¶37) comprising:
analyzing user input to determine one or more entities of a semantic framework in the user input (¶57 and Fig. 8A, step 802, receive and process natural language input; step 804, query the received input against one or more specified knowledge graphs / KG / ontology; in view of ¶49, query the KG to fetch all triplets with the same subject-entity and relationship per step 518),
determine one or more triples of the semantic framework for at least a portion of the one or more entities or for associated entities (¶57, query processing produces one or more triplets, each triplet includes a subject, object, and an associated relationship and for each triplet determined to be associated with the query input, identify the associated veracity values; see e.g., ¶50 and Table 2, Subject-Entity “George Washington”, Relationship “Born on”, Subject-Entity-Value “Feb. 22, 1732”);
adding to the user input to provide modified user input comprising
at least a portion of the one or more triples, or a representation thereof (¶59, step 840, modifying the NL input to correspond to the identified triplet);
an instruction to only consider the at least a portion of the one or more triples, or a representation thereof, as a sole factual basis for generating a response to the user input (¶37, select the triplet with the highest veracity value and determine that there is a conflict with the NL input, correct the NL input by replacing the language of the NL input with the selected triplet from the generated list; e.g., ¶51, if a triplet with a higher confidence score is realized, automatically replace the original value of the subject entity value with a value that has the higher veracity score, wherein the replacement is against the query and is not reflected in the KG; i.e., selecting the highest veracity triplet from the KG to replace corresponding original value of the natural language input is the instruction to the MLM that the original value in the natural language input is an error / conflict (¶37) and thus should not be the basis to generate final response while the replacement with the highest veracity value should be the factual / veracity basis to generate the final response); and
an instruction directing a machine learning model to treat the added triples, or the representation thereof, as system provided new information introduced by the computer system (¶37 and ¶60, determine a partial match between the natural language input and at least one of the identified triplets to create a new triplet for entry in the associated KG; ¶61, when KG / ontology (¶24, ontology is in the form of a KG with the facts or mentions represented as nodes in the graph) is modified to realize new entities and relationships, utilize the new information to automate training of the MLM to augment an existing MLM; i.e., this is an instruction to direct the MLM to treat the new triple with new entities and relationships as new information in order to augment the existing MLM).
It would’ve been obvious to one ordinarily skilled in the art before the effective filing date of the invention to add to the user input to provide modified user input to further include (2) an instruction to only consider the at least a portion of the one or more triples, or a representation thereof, as a sole factual basis for generating a response to the user input in order to resolve a conflict or error associated with the user input (Bacarella, ¶37).
Further, it would’ve been obvious to one ordinarily skilled in the art before the effective filing date of the invention to add to the user input to provide modified user input to further include (3) an instruction directing a large language model to treat the added triples, or the representation thereof, as system provided new information introduced by the computing system in order to augment an existing machine learning model / large language model (Bacarella, ¶61).
Further regarding claim 20, Bhatia discloses one or more computer-readable storage media1 comprising: computer-executable instructions that, when executed by a computing system comprising at least one hardware processor and at least one memory coupled to the at least one hardware processor, cause the computing system to implement the functions of claims 1 and 18 (Col 3, Rows 41-43, functions carried out by a processor executing instructions stored in memory).
Regarding Claim 2, Bhatia discloses wherein the analyzing the at least a portion of the plurality of tokens comprises providing the at least a plurality of tokens to a named entity recognition service (Col 4, Rows 54-61, provide natural language input 101 to named-entity recognition system 114, which locates and classifies named entities mentioned in the unstructured text into pre-defined entity classes).
Regarding Claims 3 and 19, Bhatia discloses wherein adding at least a portion of the one or more triples, or a representation thereof, to the user input to provide modified using input comprises:
submitting triples of the at least a portion of the plurality of triples to a verbalization function to provide the representation, the representation being verbalized triples (Col 6, Rows 1-5, triple verbalizer 140 verbalizes the knowledge base triples by concatenating the head entity, the relationship, and the tail to form a natural language phrase).
Regarding Claim 4, Bhatia discloses wherein the semantic framework comprises a knowledge graph (Col 5, Rows 26-30, knowledge base 120 takes the form of a knowledge graph).
Regarding Claim 5, Bhatia discloses identifying one or more associated entities for a subset of the one or more entities by traversing the semantic framework through one or more levels of indirection from each respective entity within the set of associated entities (Col 4, Rows 9-15, for “I like whales” that is not a question or query (i.e., indirect input or natural language input without explicit direction or instruction), use the knowledge base to retrieve knowledge about whales to formulate the response “did you know that blue whales are the largest whales?”; per Col 5, Rows 10-16, retrieve information from knowledge base for each entity identified by searching (i.e., traversing) the knowledge base for an identified entity; e.g., “I like whales”, retrieve information about “blue whales” with relationship labeled “largest label”).
Regarding Claim 6, Bhatia discloses wherein the identifying is carried out up to a specified level of indirection (Col 4, Rows 9-15, for “I like whales”, generating the response “did you know that blue whales are the largest whales?” requires searching for at least “blue whales” even though “I like whales” did not direct or instruct conversational system to search / query “blue whales”).
Regarding Claim 7, Bhatia discloses wherein the identifying is carried out until a threshold number of entities has been identified (Col 6, Rows 23-29 in view of Col 2, Rows 61-64 and Col 5, Rows 14-18 and Rows 30-32, information retrieved for identified entity comprises knowledge base triple where two nodes of each triple correspond to entities; select top-k triples where k is a threshold means selecting top k entities corresponding to nodes of knowledge base triples).
Regarding Claim 8, Bhatia discloses wherein the triples are in the form of (subject, object, predicate) (Col 5, Rows 19-20, a knowledge base triple comprises a subject, a predicate, and an object), and the identifying one or more associated entities is carried out for relationships where a respective entity of the one or more entities serves as a subject and for relationships where a respective entity of the one or more entities serves as an object (Col 5, Rows 14-22, retrieve triples from the knowledge base for an identified entity where the identified entity is either the subject or the object of a triple retrieved from the knowledge base).
Regarding Claim 9, Bhatia discloses wherein the modified input is not provided to the user (Col 6, Rows 25-28, select top-k triples 150 to be fed as input along with x to the NLP model 112; in view of Col 2, Rows 61-64, provide a threshold amount of triples as input to the NLP model along with a natural language input).
Regarding Claim 10, Bhatia discloses wherein the user input prior to modification is not provided to the large language model without the content of the modification (Col 2, Rows 61-64, provide a threshold amount of triples as input to NLP model along with a natural language input so that the NLP model processes both the natural language input and the triples to generate a response).
Regarding Claim 11, Bhatia discloses adding a length constraint to the modified user input (Col 2, Rows 61-64, provide a threshold amount (e.g., top five) of triples along with a natural language input; see also Col 6, Rows 29-31).
Regarding Claim 13, Bhatia discloses adding a contextual instruction to the modified user input (Col 6, Rows 25-28, select the top-k triples 150 that constitute the contextual knowledge to be fed as input along with input text x to the NLP model 112).
Regarding Claim 16, Bhatia discloses wherein the based on the analyzing, determining one or more entities of a semantic framework represented in the at least a portion of the plurality of tokens comprises analyzing multiple discrete semantic frameworks (Col 8, Rows 52-55, retrieve information related to an entity in the plurality of entities from a knowledge base; per Col 2, Rows 39-41, analyzing knowledge bases storing explicit relationships between entities to determine the knowledge base to retrieve information in the form of a triple).
Regarding Claim 17, Bhatia discloses wherein the based on the analyzing, determining one or more entities of a semantic framework represented in the at least a portion of the plurality of tokens comprises sending an analysis request to be executed on a semantic framework located on a remote computing system (Col 3, Rows 18-23, server side device in environment 100 implementing operations of embedder 112, NLP model 112, NER model 114, knowledge base 120, triple verbalizer 140, and similarity ranker 142).
Claim 15 is rejected under 35 USC 103(a) as being unpatentable over Bhatia et al. (US 11997056 B2) in view of Bacarella et al. (US 20019/0213258 A1) as applied to claim 1, in further view of Raniere (US 20015/0081663 A1).
Regarding Claim 15, Bhatia discloses wherein the one or more entities correspond to a first set of one or more entities (Col 4, Rows 9-15, for “I like whales”, search for entity such as “whales” in the knowledge base), the operations further comprising:
identifying a second set of one or more entities in the response (Col 4, Rows 9-15, for “I like whales”, generating the response “did you know that blue whales are the largest whales?” means the conversational system searched the knowledge base for “blue whales”);
linking one or more entities of the second set of one or more entities to supplemental content (Col 4, Rows 9-15, generating the response “did you know that blue whales are the largest whales?” based on entity “blue whales”),
wherein the displaying the response in response to the user input comprises displaying the response (Col 3, Rows 25-29, information such as response “did you know that blue whales are the largest whales” are displayed on a suer device).
Bhatia does not disclose wherein the displaying the response in response to the user input comprises displaying the response with one or more links to supplemental content.
Raniere discloses a computing system analyzing user input to determine one or more entities corresponding to a first set of one or more entities (¶43, based on user message that he is planning on taking a vacation to a beach, processor 103 recognizes “planning”, “vacation”, and “beach” as keywords; per ¶5, parsing user input into at least one keyword to search an information repository for information related to the keyword), identifying a second set of one or more entities in response (¶43, suggest other content such as rental cars, hotels, surfing lessons, boat rentals; per ¶51, add or identify related keywords to the list of keywords parsed from received user input), linking one or more entities of the second set of one or more entities to supplemental content and displaying a response to the user input comprises displaying the response with one or more links to supplemental content (¶43, suggest beach vacation destination and in addition other content such as rental cars, hotels, surfing lessons, boat rentals etc.; in view of ¶62, the suggestion includes presenting / suggesting the user to check out corresponding links by displaying information retrieved relating to a search of the information repository per ¶5), and receiving user input selecting a linked entity and displaying the supplemental content for the linked entity (¶27, user may manually select the information being collected pertaining to active search of user inputs; per ¶51, the step of identifying related keywords bolsters the amount of overall content researched by the computing system and ultimately presented to the user; i.e., upon user manually selecting suggested links to additional keywords “rental cars”, “hotels”, “surfing lessons”, and “boat rentals”, the system present / display the overall content corresponding to “rental cars”, “hotels”, “surfing lessons”, and “boat rentals”).
It would’ve been obvious to one ordinarily skilled in the art before the effective filing date of the invention to display the response with one or more links to supplemental content for user to select a linked entity and to display supplemental content for the linked entity in order to bolster the amount of overall content that may be researched by the computing system based on related keywords / entities and ultimately presented to the user (Raniere, ¶51).
Claim 21 is rejected under 35 USC 103(a) as being unpatentable over Bhatia et al. (US 11997056 B2) in view of Bacarella et al. (US 20019/0213258 A1) as applied to claim 1, in further view of Padmanabhan et al. (US 2025/0005299 A1).
Regarding Claim 21, Bhatia does not disclose wherein the modified user input further comprises an instruction restricting the large language model from generating content that violates a domain specific policy associated with the semantic framework.
Padmanabhan discloses a database system with a generative language model (¶19 and Fig. 1) receiving user input to determine instructions and policies to include in a prompt to the generative language model (¶20) comprising an instruction restricting the large language model from generating content that violates a domain specific policy associated with a semantic framework (¶48, determine one or more policies for the prompt template based on user input, client machine tenant, or database system service provider; ¶49, potential policies include restrictions on the type of output generated by the generative language model, restrictions on access to or use of potentially sensitive information, restrictions on the use of intellectual property and the like).
It would’ve been obvious to one ordinarily skilled in the art before the effective filing date of the invention to add, to the modified user input, an instruction restricting the large language model from generating content that violates a domain specific policy associated with the semantic framework in order to restrict the type of output generated by a generative language model / LLM (Padmanabhan, ¶49).
Conclusion
Applicant's amendment necessitated the new grounds 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 extension fee 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 examiner Richard Z. Zhu whose telephone number is 571-270-1587 or examiner’s supervisor Hai Phan whose telephone number is 571-272-6338. Examiner Richard Zhu can normally be reached on M-Th, 0730:1700.
Information regarding the status of an application may be obtained from the Patent Application Information Retrieval (PAIR) system. Status information for published applications may be obtained from either Private PAIR or Public PAIR. Status information for unpublished applications is available through Private PAIR only. For more information about the PAIR system, see http://pair-direct.uspto.gov. Should you have questions on access to the Private PAIR system, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative or access to the automated information system, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/RICHARD Z ZHU/Primary Examiner, Art Unit 2654 03/07/2026
1 Specification, US 2025/0094707 A1 at ¶374: “The term computer-readable storage media does not include signals and carrier waves”.