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 .
DETAILED ACTION
Claims 1-20 are pending.
This action is response to the application filed on November 06, 2024.
Claim Rejections - 35 USC § 102
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 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 the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(2) the claimed invention was described in a patent issued under section 151, or in an application for patent published or deemed published under section 122(b), in which the patent or application, as the case may be, names another inventor and was effectively filed before the effective filing date of the claimed invention.
Claims 1-20 are rejected under 35 U.S.C. 102(a)(2) as being anticipated by Zhang et al (US 20260119803 A1).
With respect to claims 1, 10 and 17, Zhang et al teaches
receiving, via a user interface (UI) of an application, a question based on content from a plurality of distinct data sources ([0003] language models, such as large language model(s) (LLM(s)), to perform natural language processing (NLP) tasks. A language model is a type of machine learning (ML) model that supports NLP tasks, such as generating text, analyzing sentiments, answering prompts (e.g., specific instructions and/or requests posed in natural language) in a conversational manner [0042] Client device 150(1) and client device 150(2) may each include a user interface (UI) 152(1), 152(2));
determining one or more keywords in the question using an embedding large language model (LLM) of a generative artificial intelligence (AI) system, wherein the one or more keywords are determined based on a semantic analysis of embeddings generated from the question by the embedding LLM ([0061] prompting first language model 211 to generate synthetic data 214 (e.g., questions, question-context tuples, question-answer tuples, question-context-answer tuples, etc.). A prompt may comprise an input query or instruction, such as a natural language question, a keyword search);
performing a search of the content from the plurality of data sources using the one or more keywords, wherein the search is performed in a plurality of data formats for the plurality of data sources using application programming interface (API) calls to APIs associated with search functions of the plurality of distinct data sources ([0061] prompting first language model 211 to generate synthetic data 214 (e.g., questions, question-context tuples, question-answer tuples, question-context-answer tuples, etc.). A prompt may comprise an input query or instruction, such as a natural language question, a keyword search. [0029] Depending on the model, data formats may be more effective than others for fine-tuning to better results compared to other formats (e.g., such as a raw text corpus, also referred to herein as a “raw information item”). Thus, strategies for formatting knowledge prior to LLM fine-tuning);
identifying one or more matches of the content from the plurality of distinct data sources to the question based on the search, wherein each of the one or more matches comprises data in a corresponding one of the plurality of data formats ([0033] The question-answering format of synthetic data generated via interleaved generation may naturally mirror the process of information-seeking, providing direct contextual alignment and relevance between the questions and their respective answers);
generating an answer to the question based on the one or more matches using a summarization LLM of the generative AI system, wherein the answer comprises a text generated by the summarization LLM from the plurality of data formats ([0002] artificial intelligence (AI) humans using natural language. Dialogue systems, which can communicate with users in natural language, may carry out unstructured conversations, with users, on any topic (e.g., open-domain systems). Performant dialogue systems exhibit competence in understanding natural language, making informed decisions, and generating fluent, engaging, contextually appropriate, and accurate responses. [0003] large language model(s) (LLM(s)), to perform natural language processing (NLP) tasks. A language model is a type of machine learning (ML) model that supports NLP tasks, such as generating text, analyzing sentiments, answering prompts (e.g., specific instructions and/or requests posed in natural language) in a conversational manner, translating text from one language to another, and/or the like. Language models make it possible for software to “understand” typical human speech or written content. generating human-understandable responses through natural language generation (NLG)); and
outputting, via the UI, the answer to the question [0003] large language model(s) (LLM(s)), to perform natural language processing (NLP) tasks. A language model is a type of machine learning (ML) model that supports NLP tasks, such as generating text, analyzing sentiments, answering prompts (e.g., specific instructions and/or requests posed in natural language) in a conversational manner, translating text from one language to another, and/or the like. Language models make it possible for software to “understand” typical human speech or written content. generating human-understandable responses through natural language generation).
With respect to claims 2 and 13, Zhang et al teaches natural language question, and the determining the set of keywords includes predicting an intent and determining a context for the question using a natural language processor (NLP) of the generative AI system ([0002] artificial intelligence (AI) humans using natural language. Dialogue systems, which can communicate with users in natural language, may carry out unstructured conversations, with users, on any topic (e.g., open-domain systems). Performant dialogue systems exhibit competence in understanding natural language, making informed decisions, and generating fluent, engaging, contextually appropriate, and accurate responses. [0003] large language model(s) (LLM(s)), to perform natural language processing (NLP) tasks. A language model is a type of machine learning (ML) model).
With respect to claim 3, Zhang et al teaches converting the question to a first vector using the embedding LLM; converting search results from the search to one or more second vectors using the embedding LLM; comparing the first vector to the one or more second vectors based on a vector comparison function; and determining the one or more matches from the search results based on the vector comparison function and a similarity threshold ([0054] performing one or more other segmentation techniques to divide raw information item 204 into, or otherwise extract, smaller units of information. These techniques may include paragraph-based segmentation, topic-based segmentation, such as using natural language processing techniques, semantic similarity-based clustering, named entity recognition for entity-centric segmentation, and/or temporal or chronological segmentation for time-based content).
With respect to claim 4, Zhang et al teaches extracting the one or more keywords from the question using an NLP and one or more third vectors generated for the one or more keywords by the embedding LLM, wherein the converting the question to the first vector is based, at least in part, on the one or more third vectors generated for the one or more keywords ([0054] performing one or more other segmentation techniques to divide raw information item 204 into, or otherwise extract, smaller units of information. paragraph-based segmentation, topic-based segmentation, such as using natural language processing techniques, semantic similarity-based clustering, named entity recognition for entity-centric segmentation, and/or temporal or chronological segmentation for time-based content).
With respect to claim 5, Zhang et al teaches prompting the summarization LLM with an instruction to generate the answer using the one or more matches each in the corresponding one of the plurality of data formats, and wherein the answer summarizes the one or more matches in a text format corresponding to the question ([0029] Depending on the model, data formats may be more effective than others for fine-tuning to better results compared to other formats (e.g., such as a raw text corpus, also referred to herein as a “raw information item”). Thus, strategies for formatting knowledge prior to LLM fine-tuning)
With respect to claim 6, Zhang et al teaches receiving feedback associated with the answer; and updating at least one data retrieval module associated with the performing the search based on the feedback ([0069] Fine-tuning 216 results in obtaining a fine-tuned second language model 220 (e.g., a fine-tuned version of second language model 217).
With respect to claim 7, Zhang et al teaches receiving feedback associated with the answer; and updating at least one data retrieval module associated with the performing the search based on the feedback ([0069] Fine-tuning 216 results in obtaining a fine-tuned second language model 220 (e.g., a fine-tuned version of second language model).
With respect to claim 8, Zhang et al teaches matches ranked based on a relevancy score of each of the one or more matches to the question, and wherein the summarization and the one or more matches ranked are provided via the UI for the answer ([0069] Fine-tuning 216 results in obtaining a fine-tuned second language model 220 (e.g., a fine-tuned version of second language model).
With respect to claims 9 and 20, Zhang et al teaches at least one of an internal chat platform, a service ticketing platform, a code collaboration workspace platform, or computing service documentation ([0069] Fine-tuning 216 results in obtaining a fine-tuned second language model 220 (e.g., a fine-tuned version of second language model 217).
With respect to claim 11, Zhang et al teaches links in the text to the content from the plurality of data sources ([0038] FIG. 1 system 100 supporting a microservice 104(1) (e.g., software-defined service, which in some cases, may be cloud-native) implementing one or more language models 108, such as LLM(s)).
With respect to claim 12, Zhang et al teaches citations in corresponding portions of the text to the content ([0038] FIG. 1 system 100 supporting a microservice 104(1) (e.g., software-defined service, which in some cases, may be cloud-native) implementing one or more language models 108, such as LLM(s)).
With respect to claim 13, Zhang et al teaches predicting an intent and determining a context for the question using a natural language processor (NLP) of the generative AI system ([0038] FIG. 1 system 100 supporting a microservice 104(1) (e.g., software-defined service, which in some cases, may be cloud-native) implementing one or more language models 108, LLM(s)).
With respect to claim 14, Zhang et al teaches API call to the API utilizes a search function associated with the API and the one of the data formats to search the corresponding one of the plurality of data sources ([0038] FIG. 1 system 100 supporting a microservice 104(1) (e.g., software-defined service, which in some cases, may be cloud-native) implementing one or more language models 108, such as LLM(s)).
With respect to claim 15, Zhang et al teaches receive an additional question that requests one of a refinement of the answer or additional information associated with the content used for the text in the answer, wherein the additional question is associated with the question previously asked; determine a change to the content based on the additional question; and update the answer using the summarization LLM and based on the change to the content ([0002] artificial intelligence (AI) is to create machines capable of understanding and engaging in conversation with humans using natural language. Dialogue systems, which can communicate with users in natural language, may carry out unstructured conversations, with users, on any topic (e.g., open-domain systems)).
With respect to claim 16, Zhang et al teaches additional question is received via a user interface field provided with the answer for the refinement or the additional information ([0002] artificial intelligence (AI) is to create machines capable of understanding and engaging in conversation with humans using natural language. Dialogue systems, which can communicate with users in natural language, may carry out unstructured conversations, with users, on any topic (e.g., open-domain systems)).
With respect to claim 18, Zhang et al teaches generating the set of keywords from the question using an embedding LLM of the generative AI system, wherein the set of keywords are determined based on a semantic analysis of embeddings generated from the question by the embedding LLM ([0002] artificial intelligence (AI) is to create machines capable of understanding and engaging in conversation with humans using natural language).
With respect to claim 19, Zhang et al teaches answer is requested to be provided in natural language as a summarization of the content in place of search results via the UI ([0002] artificial intelligence (AI) is to create machines capable of understanding and engaging in conversation with humans using natural language).
Conclusion
The prior arty made of record and not relied upon is considered pertinent to applicant’s disclosure.
Fieldman (US 12265788 B1) Systems And Methods For Connected Natural Language Models
Considered for teaching generally for providing answer data through multiple connected large language models. Operations may include receiving, through a graphical user interface associated with a local large language model having access to a first limited private dataset but not a second limited private dataset, an input from a user device, identifying, based on the input, an external large language model from among a plurality of external large language models, transmitting the input to the external large language model, receiving, from the external large language model, the answer data responsive to the input, generating, by the local large language model, response data based on the answer data, and outputting the response data at the user device.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to ISAAC M WOO whose telephone number is (571)272-4043. The examiner can normally be reached 9:00 to 5:00.
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, Tony Mahmoudi can be reached at 571-272-4078. 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.
/ISAAC M WOO/ Primary Examiner, Art Unit 2163
Prosecution Insights