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
A filing date of 05/07/2024 is acknowledged.
Claims 1 – 20 are pending.
Claim Objections
Claims 1 – 20 are objected to because of the following informalities:
Claim 1
Line 9; change “the use” to --a use--.
Claim 2
Line 3; change “the project” to --a project--.
Claims 3 – 6
These claims are dependent claims of claim 1 either directly or indirectly; therefore, they inherit issue(s) of claim 1.
Claim 7
Line 1; insert --the-- before “multiple chunks”.
Claims 8 – 12
These claims are dependent claims of claim 1 either directly or indirectly; therefore, they inherit issue(s) of claim 1.
Claim 13
Line 3; insert --the-- before “at least two example first chunks”.
Claim 14
The claim is dependent claims of claim 1; therefore, it inherits issue(s) of claim 1.
Claim 15
Line 1; change “a hallucination” to --the hallucination--.
Line 3; change “the use” to --a use--.
Line 5; change “a hallucination” to --the hallucination--. And, insert --the-- before “information”.
Claim 16
The claim is dependent claims of claim 15; therefore, it inherits issue(s) of claim 15.
Claim 17
Line 2; insert --the-- before “at least one hallucination”.
Line 3, “the process” lacks proper antecedent basis.
Claim 18
Line 2; insert --the-- before “at least one hallucination”.
Claim 19
Lines 2 and 3; insert --the-- before “at least one hallucination” respectively.
Claim 20
The claim is dependent claims of claim 1; therefore, it inherits issue(s) of claim 1.
Appropriate correction is required.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 1 – 20 are rejected under 35 U.S.C. 101 because the claimed invention is directed to a judicial exception (i.e., a law of nature, a natural phenomenon, or an abstract idea) without significantly more.
Claim 1
Step 1
The claim is statutory because it is directed to a method.
Step 2A, prong 1
The claim recites limitations:
“dependent upon the topical information, determining a first chunk of text to generate;
retrieving, from an index, at least one example first chunk of text;
generating, by the first large language model, the first chunk of text dependent upon the topical information and accomplishing the desired purpose set out in the prompt.”
The limitation “dependent upon the topical information, determining a first chunk of text to generate” is classified as abstract idea because it relies on human observation of topical information to decide whether a first chunk of text is to be generated.
Similarly, the limitation “retrieving, from an index, at least one example first chunk of text” is also classified as abstract idea of mental process as it involves human evaluation of index to retrieve at least one example chunk of text. Furthermore, the limitation “generating … the first chunk of text dependent upon the topical information and accomplishing the desired purpose set out in the prompt” can be performed by human to generate the chunk of text based on topical information and desired purpose with an aid of paper and pen.
These limitations are not considered as an improvement and, therefore, are not integrated with any additional element(s) that integrate the judicial exception into a practical application. These limitations are directed to a mental process.
Step 2A, prong 2
The claim recites additional elements “a computer processor and a first large language model.” The additional elements are recited as high level of generality and used as a tool to perform the limitations. Thus, the additional elements are not indicative of an integration into a practical application.
The claim further recites addiction limitations:
“receiving topical information relevant to the document to be created;
providing the at least one example first chunk of text and at least a portion of the topical information to a first large language model;
prompting … the first large language model to generate the first chunk of text through the use of a first request that includes a prompt that states a desired purpose of the first chunk of text …, a context …, and the at least one example first chuck of text.”
These additional limitations merely receive information, provide chunks of text and topical information, and prompt the first large language model to generate the first chunk of text. These actions are insignificant extra-solution activity. Thus, they do not integrate the judicial exception into a practical application.
Steps 2B
The claim as a whole is not amounted to significantly more than the judicial exception. Claim 7 is directed to an abstract idea and, therefore, is not patent eligible.
Analysis of claims 2 – 20
Claim 2
The claim recites “the topical information includes at least one of the following: a project name, a project identification number, a client name, a client industry, a client description, a document type, at least one challenge of the project, a project duration, at least one priority of the project, at least one special consideration, at least one service type, a delivery type, and a delivery location.”
The limitation defines topical information. Thus, the limitation is an insignificant extra-solution activity and it is not integrated into a practical application because they do not impose any meaningful limits on practicing the abstract idea. So, it does not include any additional element that is sufficient to amount to significantly more than the judicial exception.
Claim 3
The claim recites “the document is a contract.”
The limitation defines the document. Thus, the limitation is an insignificant extra-solution activity and it is not integrated into a practical application because they do not impose any meaningful limits on practicing the abstract idea. So, it does not include any additional element that is sufficient to amount to significantly more than the judicial exception.
Claim 4
The claim recites “the contract is a statement of work.”
The limitation defines the contract. Thus, the limitation is an insignificant extra-solution activity and it is not integrated into a practical application because they do not impose any meaningful limits on practicing the abstract idea. So, it does not include any additional element that is sufficient to amount to significantly more than the judicial exception.
Claim 5
The claim recites “the statement of work is for development of a software program for a client.”
The limitation defines the statement of work. Thus, the limitation is an insignificant extra-solution activity and it is not integrated into a practical application because they do not impose any meaningful limits on practicing the abstract idea. So, it does not include any additional element that is sufficient to amount to significantly more than the judicial exception.
Claim 6
The claim recites “the desired purpose of the first chunk for the statement of work is at least one of the following: a project scope, a project summary, an executive summary, client responsibilities, a project description, deliverables, assumptions, a project duration, a service description, and party roles.”
The limitation defines the desired purpose. Thus, the limitation is an insignificant extra-solution activity and it is not integrated into a practical application because they do not impose any meaningful limits on practicing the abstract idea. So, it does not include any additional element that is sufficient to amount to significantly more than the judicial exception.
Claim 7
The claim recites “the statement of work includes multiple chunks, one of which is the first chunk.”
The limitation defines the statement of work. Thus, the limitation is an insignificant extra-solution activity and it is not integrated into a practical application because they do not impose any meaningful limits on practicing the abstract idea. So, it does not include any additional element that is sufficient to amount to significantly more than the judicial exception.
Claim 8
The claim recites “assembling the multiple chunks into one continuous document.”
The limitation assembles multiple chunks. Thus, the limitation is an insignificant extra-solution activity and it is not integrated into a practical application because they do not impose any meaningful limits on practicing the abstract idea. So, it does not include any additional element that is sufficient to amount to significantly more than the judicial exception.
Claim 9
The claim recites “the multiple chunks are separate from one another within the document by corresponding headings.”
The limitation arranges the multiple chunks. Thus, the limitation is an insignificant extra-solution activity and it is not integrated into a practical application because they do not impose any meaningful limits on practicing the abstract idea. So, it does not include any additional element that is sufficient to amount to significantly more than the judicial exception.
Claim 10
The claim recites “the index includes multiple example first chunks; and searching the index … for the at least one example first chunk.”
The limitation “the index includes multiple example first chunks” defines the index arranges the multiple chunks. And, the limitation “searching the index … for the at least one example first chunk” merely searches index. Thus, the limitations are insignificant extra-solution activities. They are not integrated into a practical application because they do not impose any meaningful limits on practicing the abstract idea. So, it does not include any additional element that is sufficient to amount to significantly more than the judicial exception.
Claim 11
The claim recites “the searching is performed via a similarity search.”
The limitation indicates method of searching. Thus, the limitations are insignificant extra-solution activities. They are not integrated into a practical application because they do not impose any meaningful limits on practicing the abstract idea. So, it does not include any additional element that is sufficient to amount to significantly more than the judicial exception.
Claim 12
The claim recites “the searching of the index is dependent upon the topical information so that the at least one example first chunk is relevant to the first chunk to be generated.”
The limitation indicates method of searching. Thus, the limitations are insignificant extra-solution activities. They are not integrated into a practical application because they do not impose any meaningful limits on practicing the abstract idea. So, it does not include any additional element that is sufficient to amount to significantly more than the judicial exception.
Claim 13
The claim recites “the at least one example first chunk includes at least two example first chunks relevant to the first chunk to be generated such that the step of prompting the first large language model includes providing at least two example first chunks of text to the first large language model.”
The limitation defines the at least one example first chunk. Thus, the limitations are insignificant extra-solution activities. They are not integrated into a practical application because they do not impose any meaningful limits on practicing the abstract idea. So, it does not include any additional element that is sufficient to amount to significantly more than the judicial exception.
Claim 14
The claim recites “evaluating the first chunk of text for a hallucination as generated by the first large language model.”
The limitation evaluates the first chunk for hallucination. The step of evaluation relies on human observation and evaluation of the chunk of text for hallucination. The limitation, under its broadest reasonable interpretation, covers performance of the limitation in the mind, and it is not integrated into a practical application because they do not impose any meaningful limits on practicing the abstract idea. So, it does not include any additional element that is sufficient to amount to significantly more than the judicial exception.
Claim 15
The claim recites “… to review the first chunk for a hallucination and a context that provides the first chunk and information dependent upon the topical information; and
determining … whether the first chunk of text includes at least one hallucination that includes information inconsistent with the topical information or the desired purpose of the first chunk.”
The limitation reviews and evaluates the first chunk for hallucination. The step of evaluation relies on human observation and evaluation of the chunk of text for hallucination. The limitation, under its broadest reasonable interpretation, covers performance of the limitation in the mind, and it is not integrated into a practical application because they do not impose any meaningful limits on practicing the abstract idea. So, it does not include any additional element that is sufficient to amount to significantly more than the judicial exception.
Claim 16
The claim recites “the second large language model is different from the first large language model.”
The limitation recognizes the difference between large language models. Thus, the limitations are insignificant extra-solution activities. They are not integrated into a practical application because they do not impose any meaningful limits on practicing the abstract idea. So, it does not include any additional element that is sufficient to amount to significantly more than the judicial exception.
Claim 17
The claim recites “in response to … determining that the first chunk includes at least one hallucination …
discarding the first chunk and repeating the process to generate a new first chunk.”
The limitation removes the first chunk. Thus, the limitations are insignificant extra-solution activities. They are not integrated into a practical application because they do not impose any meaningful limits on practicing the abstract idea. So, it does not include any additional element that is sufficient to amount to significantly more than the judicial exception.
Claim 18
The claim recites “in response to determining that the first chunk includes at least one hallucination, … saving the first chunk in storage media.”
The limitation saves the first chunk. Thus, the limitations are insignificant extra-solution activities. They are not integrated into a practical application because they do not impose any meaningful limits on practicing the abstract idea. So, it does not include any additional element that is sufficient to amount to significantly more than the judicial exception.
Claim 19
The claim recites “in response … determining that the first chunk includes at least one hallucination … initiating an alert that the first chunk includes at least one hallucination.”
The limitation initiates an alert. Thus, the limitations are insignificant extra-solution activities. They are not integrated into a practical application because they do not impose any meaningful limits on practicing the abstract idea. So, it does not include any additional element that is sufficient to amount to significantly more than the judicial exception.
Claim 20
The claim recites “before prompting the first large language model, assembling the first request …; and
providing the first request to the first large language model by the prompt module.”
The limitation assembles request and provides it to the large language model. Thus, the limitations are insignificant extra-solution activities. They are not integrated into a practical application because they do not impose any meaningful limits on practicing the abstract idea. So, it does not include any additional element that is sufficient to amount to significantly more than the judicial exception.
Double Patenting
The nonstatutory double patenting rejection is based on a judicially created doctrine grounded in public policy (a policy reflected in the statute) so as to prevent the unjustified or improper timewise extension of the “right to exclude” granted by a patent and to prevent possible harassment by multiple assignees. A nonstatutory double patenting rejection is appropriate where the conflicting claims are not identical, but at least one examined application claim is not patentably distinct from the reference claim(s) because the examined application claim is either anticipated by, or would have been obvious over, the reference claim(s). See, e.g., In re Berg, 140 F.3d 1428, 46 USPQ2d 1226 (Fed. Cir. 1998); In re Goodman, 11 F.3d 1046, 29 USPQ2d 2010 (Fed. Cir. 1993); In re Longi, 759 F.2d 887, 225 USPQ 645 (Fed. Cir. 1985); In re Van Ornum, 686 F.2d 937, 214 USPQ 761 (CCPA 1982); In re Vogel, 422 F.2d 438, 164 USPQ 619 (CCPA 1970); In re Thorington, 418 F.2d 528, 163 USPQ 644 (CCPA 1969).
A timely filed terminal disclaimer in compliance with 37 CFR 1.321(c) or 1.321(d) may be used to overcome an actual or provisional rejection based on nonstatutory double patenting provided the reference application or patent either is shown to be commonly owned with the examined application, or claims an invention made as a result of activities undertaken within the scope of a joint research agreement. See MPEP § 717.02 for applications subject to examination under the first inventor to file provisions of the AIA as explained in MPEP § 2159. See MPEP § 2146 et seq. for applications not subject to examination under the first inventor to file provisions of the AIA . A terminal disclaimer must be signed in compliance with 37 CFR 1.321(b).
The filing of a terminal disclaimer by itself is not a complete reply to a nonstatutory double patenting (NSDP) rejection. A complete reply requires that the terminal disclaimer be accompanied by a reply requesting reconsideration of the prior Office action. Even where the NSDP rejection is provisional the reply must be complete. See MPEP § 804, subsection I.B.1. For a reply to a non-final Office action, see 37 CFR 1.111(a). For a reply to final Office action, see 37 CFR 1.113(c). A request for reconsideration while not provided for in 37 CFR 1.113(c) may be filed after final for consideration. See MPEP §§ 706.07(e) and 714.13.
The USPTO Internet website contains terminal disclaimer forms which may be used. Please visit www.uspto.gov/patent/patents-forms. The actual filing date of the application in which the form is filed determines what form (e.g., PTO/SB/25, PTO/SB/26, PTO/AIA /25, or PTO/AIA /26) should be used. A web-based eTerminal Disclaimer may be filled out completely online using web-screens. An eTerminal Disclaimer that meets all requirements is auto-processed and approved immediately upon submission. For more information about eTerminal Disclaimers, refer to www.uspto.gov/patents/apply/applying-online/eterminal-disclaimer.
Claims 1 – 6, 14, and 16 are provisionally rejected on the ground of nonstatutory double patenting as being unpatentable over claims 1 – 6, 9, and 13 of copending Application No. 18/657,022 (reference application. Hereinafter ‘022). Although the claims at issue are not identical, they are not patentably distinct from each other.
This is a provisional nonstatutory double patenting rejection because the patentably indistinct claims have not in fact been patented.
Instant Application 18/657,040
Application 18/657,022
Claim 1
A method of generating a document having multiple chunks of text that collectively form at least a portion of the document, the method comprising:
receiving topical information relevant to the document to be created;
dependent upon the topical information, determining a first chunk of text to generate;
retrieving, from an index, at least one example first chunk of text;
providing the at least one example first chunk of text and at least a portion of the topical information to a first large language model;
prompting, by a computer processor, the first large language model to generate the first chunk of text through the use of a first request that includes a prompt that states a desired purpose of the first chunk of text to be generated, a context that provides information dependent upon the topical information, and the at least one example first chuck of text; and
generating, by the first large language model, the first chunk of text dependent upon the topical information and accomplishing the desired purpose set out in the prompt.
Claim 1
A method of generating a document having multiple chunks of text that collectively form at least a portion of the document, the method comprising:
determining a first chunk of the multiple chunks of text to generate dependent upon topical information relevant to the document that is to be created;
retrieving, from an index, at least one example first chunk of text with the at least one example first chunk of text being dependent upon a desired purpose of the first chunk and upon the topical information;
generating the first chunk of text by a first large language model via a first request that includes a prompt that states the desired purpose of the first chunk of text to be generated, a context that provides first information dependent upon the topical information, and the at least one example first chunk of text;
determining a second chunk of the multiple chunks of text to generate dependent upon the topical information;
retrieving, from the index, at least one example second chunk of text with the at least one example second chunk of text being dependent upon a desired purpose of the second chunk and upon the topical information;
generating the second chunk of text by the first large language model via a second request that includes a prompt that states the desired purpose of the second chunk of text to be generated, a context that provides second information dependent upon the topical information, the at least one example second chunk of text, and the first chunk of text with the second chunk of text being dependent upon the first chunk of text previously generated by the first large language model; and
assembling the first chunk of text and the second chunk of text to form at least a portion of the document such that the first chunk and the second chunk are consistent in content.
Claim 2
The method of claim 1, wherein the topical information includes at least one of the following: a project name, a project identification number, a client name, a client industry, a client description, a document type, at least one challenge of the project, a project duration, at least one priority of the project, at least one special consideration, at least one service type, a delivery type, and a delivery location.
Claim 2
The method of claim 1, wherein the topical information includes at least one of the following: a project name, a project identification number, a client name, a client industry, a client description, a document type, at least one challenge of the project, a project duration, at least one priority of the project, at least one special consideration, at least one service type, a delivery type, and a delivery location.
Claim 3
The method of claim 1, wherein the document is a contract.
Claim 3
The method of claim 1, wherein the document is a contract.
Claim 4
The method of claim 3, wherein the contract is a statement of work.
Claim 4
The method of claim 3, wherein the contract is a statement of work.
Claim 5
The method of claim 4, wherein the statement of work is for development of a software program for a client.
Claim 5
The method of claim 4, wherein the statement of work is for development of a software program for a client.
Claim 6
The method of claim 5, wherein the desired purpose of the first chunk for the statement of work is at least one of the following: a project scope, a project summary, an executive summary, client responsibilities, a project description, deliverables, assumptions, a project duration, a service description, and party roles.
Claim 6
The method of claim 5, wherein the desired purpose of the first chunk for the statement of work is at least one of the following: a project scope, a project summary, an executive summary, client responsibilities, a project description, deliverables, assumptions, a project duration, a service description, and party roles.
Claim 14
The method of claim 1, further comprising:
evaluating the first chunk of text for a hallucination as generated by the first large language model.
Claim 9
The method of claim 1, further comprising:
evaluating the first chunk of text for a hallucination as generated by the first large language model.
Claim 16
The method of claim 15, wherein the second large language model is different from the first large language model.
Claim 13
The method of claim 9, wherein the second large language model is different from the first large language model.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claims 1 – 20 are rejected under 35 U.S.C. 103 as being unpatentable over Arunachalam et al. (US 20250211549 A1, hereinafter Arunachalam) in view of Crouch et a. (US 20160078102 A1, hereinafter Crouch.)
Claim 1
Arunachalam teaches a method of generating a document having multiple chunks of text that collectively form at least a portion of the document (Arunachalam; abstract; The embodiments are directed to a generative artificial intelligence (AI) system for generating answers (document) to questions or generating summaries (document) from data included in various data sources and in multiple domains), the method comprising:
receiving topical information relevant to the document to be created (Arunachalam; Fig. 5; [0064 – 0066] In some embodiments, generative AI system 108 may include an information request router 504. Information request router 504 may route the information request for request-specific processing based on a type of the information request. Example types of an information request may be a question (topical information) and a request for a summary (topical information) …
[0068] One of LLM models, such as LLM 110C may receive one or more chunks in the subset of chunks to generate an answer to the question in the information request …; also see Fig. 7 & [0083 – 0091] a method for generating answer to question
[0071] … Chunk retrieving module 510 may parse the metadata tags in dictionary 410A associated with project data 202A to determine tags that indicate that the corresponding chunk may include information that contributes to the summary …; also see Fig. 8 & [0092 – 0098]: a method for generating response to information request);
dependent upon the topical information, determining a first chunk of text to generate (Arunachalam; Fig. 5; [0066 – 0068] If information request router 504 determines that the information request is a question/answer request, information request router 504 passes the information request to vector retrieving module 506 …
[0068] One of LLM models, such as LLM 110C may receive one or more chunks in the subset of chunks to generate an answer to the question in the information request. In some embodiments, LLM 110C may receive the chunk that corresponds to the most similar vector first, and then refine the answer with each subsequent chunk …; also see Fig. 7 & [0083 – 0091] a method for generating answer to question
[0071] If information request router 504 determines that the information request is a request for a summary, information request router 504 passes the information request to chunk retrieving module 510. Chunk retrieving module 510 may parse the metadata tags in dictionary 410A associated with project data 202A to determine tags that indicate that the corresponding chunk may include information that contributes to the summary…; also see Fig. 8 & [0092 – 0098]: a method for generating response to information request);
retrieving, from an index, at least one example first chunk of text (Arunachalam; [0055 – 0056] Data processing engine 114 may also generate a metadata tags 406A, and associate each metadata tag (index) in metadata tags 406A with each chunk in chunks 404A…
… From chunks 404A, LLM 110 may generate embedding vectors or simply vectors 408A … Typically, one vector (index) in vectors 408A may be generated for one chunk in chunks 404A.
Fig. 5; [0066 – 0068] … Vector retrieving module 506 may use one of LLMs 110, such as LLM 110B to generate an embedding vector from the information request …
One of LLM models, such as LLM 110C may receive one or more chunks in the subset of chunks to generate an answer to the question in the information request. In some embodiments, LLM 110C may receive the chunk that corresponds to the most similar vector first, and then refine the answer with each subsequent chunk …; also see Fig. 7 & [0083 – 0091] a method for generating answer to question
[0071] … Chunk retrieving module 510 may parse the metadata tags in dictionary 410A associated with project data 202A to determine tags that indicate that the corresponding chunk may include information that contributes to the summary…; also see Fig. 8 & [0092 – 0098]: a method for generating response to information request);
providing the at least one example first chunk of text and at least a portion of the topical information to a first large language model (Arunachalam; Fig. 1; [0031] As discussed above, generative AI system 108 includes one or more large language models (LLMs) 110 …
Fig. 5; [0066 – 0068] … Vector retrieving module 506 may use one of LLMs 110, such as LLM 110B to generate an embedding vector from the information request …
One of LLM models, such as LLM 110C may receive one or more chunks in the subset of chunks to generate an answer to the question in the information request. In some embodiments, LLM 110C may receive the chunk that corresponds to the most similar vector first, and then refine the answer with each subsequent chunk …; also see Fig. 7 & [0083 – 0091] a method for generating answer to question
[0071] … Chunk retrieving module 510 may parse the metadata tags in dictionary 410A associated with project data 202A to determine tags that indicate that the corresponding chunk may include information that contributes to the summary. Once chunk retrieving module 510 identifies a subset of chunks from chunks 404A that may contribute to the summary, chunk retrieving module 510 may forward the chunks to LLM 110C. LLM 110C may receive and summarize the subset of chunks into a summary …; also see Fig. 8 & [0092 – 0098]: a method for generating response to information request);
by a computer processor, the first large language model to generate the first chunk of text through the use of a first request a prompt (Arunachalam; Fig. 1; [0034] … If so, the chat service 112 may convert the standard or predefined commands into one of the predefined prompts in an attempt to generate more accurate responses, and transmit the predefined prompts to LLMs 110 …) a desired purpose of the first chunk of text to be generated, a context that provides information dependent upon the topical information, and the at least one example first chuck of text (Arunachalam; Fig. 5; [0066 – 0068] If information request router 504 determines that the information request is a question/answer request, information request router 504 passes the information request to vector retrieving module 506
One of LLM models, such as LLM 110C may receive one or more chunks in the subset of chunks to generate an answer to the question in the information request … One of LLM models, such as LLM 110C may receive one or more chunks in the subset of chunks to generate an answer to the question in the information request …; also see Fig. 7 & [0083 – 0091] a method for generating answer to question
[0071] If information request router 504 determines that the information request is a request for a summary, information request router 504 passes the information request to chunk retrieving module 510. Chunk retrieving module 510 may parse the metadata tags in dictionary 410A associated with project data 202A to determine tags that indicate that the corresponding chunk may include information that contributes to the summary …; also see Fig. 8 & [0092 – 0098]: a method for generating response to information request); and
generating, by the first large language model, the first chunk of text dependent upon the topical information and accomplishing the desired purpose set out in the prompt (Arunachalam; Fig. 5; [0068] One of LLM models, such as LLM 110C may receive one or more chunks in the subset of chunks to generate an answer to the question in the information request…
[0070] Once LLM 110C generates an answer, generative AI system 108 may transmit the answer in a response to generative AI chatbot interface 116. Also see Fig. 7 & [0083 – 0091] a method for generating answer to question
[0071] If information request router 504 determines that the information request is a request for a summary, information request router 504 passes the information request to chunk retrieving module 510 …
Once LLM 110C generates a summary, generative AI system 108 may transmit the summary in a response to session 502 conducted using generative AI chatbot interface 116. Also see Fig. 8 & [0092 – 0098]: a method for generating response to information request.)
But Arunachalam does not explicitly teach prompting, by a computer processor, the first large language model to generate the first chunk of text through the use of a first request that includes a prompt that states a desired purpose of the first chunk of text to be generated.
However, Crouch teaches prompting, by a computer processor, the first large language model to generate the first chunk of text through the use of a first request that includes a prompt that states a desired purpose of the first chunk of text to be generated (Crouch; Fig. 5; [0075 – 0076] In step 506, the computing device may receive a prompt requiring passage search to be performed …
In step 508, the computing device may generate a search query from the received prompt for passage search. The computing device may normalize all received prompts into a search query composed of logical operators, search keywords, concept identifiers in a format (desired purpose) that is used to conduct passage (first chunk) searches by the QA system …, Also see [0079].)
Arunachalam and Crouch are in the same analogous art as they are in the same field of endeavor, searching for documents. Therefore, it would have been obvious to a person of ordinary skill, in the art before the effective filing date of the claimed invention, to incorporate Crouch into Arunachalam to generate a prompt that states desired purpose for a search of first chunk of text, as suggested by Crouch ([0075 – 0076]), to arrive at the invention.
Claim 2
Arunachalam also teaches the topical information includes at least one of the following: a project identification number, a document type (Arunachalam; Fig. 5; [0065 – 0066] In some embodiments, LLM 110A may also identify a project identifier for a project from the dialogue or from the information request.
If information request router 504 determines that the information request is a question/answer request, information request router 504 passes the information request to vector retrieving module 506…; answer == document type
[0071] If information request router 504 determines that the information request is a request for a summary (document type), information request router 504 passes the information request to chunk retrieving module 510 …), .
Claim 3
Arunachalam also teaches the document is a contract (Arunachalam; [0027] … The requirements stage may generate documents pertaining to functional requirements, technical requirements, review and approval documentation, and/or a statement of work …)
Claim 4
Arunachalam also teaches the contract is a statement of work (Arunachalam; [0027] … The requirements stage may generate documents pertaining to functional requirements, technical requirements, review and approval documentation, and/or a statement of work …)
Claim 5
Arunachalam also teaches the statement of work is for development of a software program for a client (Arunachalam; [0019] The generative AI system may be particularly useful in summarizing projects and generating answers to questions associated with projects. Projects, such as software projects, may involve various stages of development, including a planning stage, a development stage, a testing and quality assurance stage, and a release stage …)
Claim 6
Arunachalam also teaches the desired purpose of the first chunk for the statement of work is at least one of the following: a project scope, a project summary, deliverables,
Claim 7
Arunachalam also teaches the statement of work includes multiple chunks, one of which is the first chunk (Arunachalam; [0042 – 0044] During an ingestion stage, data processing engine 114 may receive project data 202A. Data processing engine 114 may convert project data 202A, including documents 204A and transcripts 206A into a uniform data 205A … In this case, data processing engine 114 may divide uniform data 205A into chunks that are less than the predefined input size, and the instance of LLM 110 may receive uniform data 205A as chunks …
Fig. 5; [0068] One of LLM models, such as LLM 110C may receive one or more chunks in the subset of chunks to generate an answer to the question in the information request …
[0071] … Once chunk retrieving module 510 identifies a subset of chunks from chunks 404A that may contribute to the summary, chunk retrieving module 510 may forward the chunks to LLM 110C. LLM 110C may receive and summarize the subset of chunks into a summary …)
Claim 8
Arunachalam also teaches assembling the multiple chunks into one continuous document (Arunachalam; Fig. 5; [0068] … Next, LLM 110C may receive a second chunk and the answer generated from the first chunk to generate a second answer …
[0071] LLM 110C may receive and summarize the subset of chunks into a summary …)
Claim 10
Arunachalam also teaches searching the index, by a search engine, for the at least one example first chunk (Arunachalam; [0055 – 0056] Data processing engine 114 may also generate a metadata tags 406A, and associate each metadata tag (index) in metadata tags 406A with each chunk in chunks 404A…
… From chunks 404A, LLM 110 may generate embedding vectors or simply vectors 408A … Typically, one vector (index) in vectors 408A may be generated for one chunk in chunks 404A.
Fig. 5; [0066 – 0068] … Vector retrieving module 506 may use one of LLMs 110, such as LLM 110B to generate an embedding vector from the information request …
One of LLM models, such as LLM 110C may receive one or more chunks in the subset of chunks to generate an answer to the question in the information request. In some embodiments, LLM 110C may receive the chunk that corresponds to the most similar vector first, and then refine the answer with each subsequent chunk …; also see Fig. 7 & [0083 – 0091] a method for generating answer to question
[0071] … Chunk retrieving module 510 may parse the metadata tags in dictionary 410A associated with project data 202A to determine tags that indicate that the corresponding chunk may include information that contributes to the summary…; also see Fig. 8 & [0092 – 0098]: a method for generating response to information request)
Claim 11
Arunachalam also teaches the searching is performed via a similarity search (Arunachalam; [0055 – 0056] Data processing engine 114 may also generate a metadata tags 406A, and associate each metadata tag (index) in metadata tags 406A with each chunk in chunks 404A…
… From chunks 404A, LLM 110 may generate embedding vectors or simply vectors 408A … Typically, one vector (index) in vectors 408A may be generated for one chunk in chunks 404A.
Fig. 5; [0066 – 0068] … Vector retrieving module 506 may use one of LLMs 110, such as LLM 110B to generate an embedding vector from the information request …
One of LLM models, such as LLM 110C may receive one or more chunks in the subset of chunks to generate an answer to the question in the information request. In some embodiments, LLM 110C may receive the chunk that corresponds to the most similar vector first, and then refine the answer with each subsequent chunk …; also see Fig. 7 & [0083 – 0091] a method for generating answer to question
[0071] … Chunk retrieving module 510 may parse the metadata tags in dictionary 410A associated with project data 202A to determine tags that indicate that the corresponding chunk may include information that contributes to the summary…; also see Fig. 8 & [0092 – 0098]: a method for generating response to information request)
Claim 12
Arunachalam also teaches the searching of the index is dependent upon the topical information so that the at least one example first chunk is relevant to the first chunk to be generated (Arunachalam; [0055 – 0056] Data processing engine 114 may also generate a metadata tags 406A, and associate each metadata tag (index) in metadata tags 406A with each chunk in chunks 404A…
… From chunks 404A, LLM 110 may generate embedding vectors or simply vectors 408A … Typically, one vector (index) in vectors 408A may be generated for one chunk in chunks 404A.
Fig. 5; [0066 – 0068] … Vector retrieving module 506 may use one of LLMs 110, such as LLM 110B to generate an embedding vector from the information request …
One of LLM models, such as LLM 110C may receive one or more chunks in the subset of chunks to generate an answer to the question in the information request. In some embodiments, LLM 110C may receive the chunk that corresponds to the most similar vector first, and then refine the answer with each subsequent chunk …; also see Fig. 7 & [0083 – 0091] a method for generating answer to question
[0071] … Chunk retrieving module 510 may parse the metadata tags in dictionary 410A associated with project data 202A to determine tags that indicate that the corresponding chunk may include information that contributes to the summary…; also see Fig. 8 & [0092 – 0098]: a method for generating response to information request)
Claim 13
Arunachalam also teaches the at least one example first chunk includes at least two example first chunks relevant to the first chunk to be generated such that the step of prompting the first large language model includes providing at least two example first chunks of text to the first large language model (Arunachalam; Fig. 5; [0066 – 0068] … Vector retrieving module 506 may use one of LLMs 110, such as LLM 110B to generate an embedding vector from the information request …
One of LLM models, such as LLM 110C may receive one or more chunks in the subset of chunks to generate an answer to the question in the information request. In some embodiments, LLM 110C may receive the chunk that corresponds to the most similar vector first, and then refine the answer with each subsequent chunk …; also see Fig. 7 & [0083 – 0091] a method for generating answer to question
[0071] … Chunk retrieving module 510 may parse the metadata tags in dictionary 410A associated with project data 202A to determine tags that indicate that the corresponding chunk may include information that contributes to the summary. Once chunk retrieving module 510 identifies a subset of chunks from chunks 404A that may contribute to the summary, chunk retrieving module 510 may forward the chunks to LLM 110C. LLM 110C may receive and summarize the subset of chunks into a summary…; also see Fig. 8 & [0092 – 0098]: a method for generating response to information request.) (Emphasis added.)
Claim 14
Arunachalam also teaches evaluating the first chunk of text for a hallucination as generated by the first large language model (Arunachalam; [0076] In some embodiments, generative AI system 108 may reduce or eliminate a number of AI hallucinations from the response. An AI hallucination may occur when one of LLMs 110 may create an answer or a summary that is not based on chunks 404. To reduce the likelihood of the hallucinations, the response displayed in session 502 of generative AI chatbot interface 116 may also include chunks 404 and/or original documents 204 and/or transcripts 206, or links to chunks 404 and/or original documents 204 and/or transcripts 206 from which LLMs 110 generated a response.)
Claim 15
Arunachalam also teaches prompting a second large language model through the use of a second request that includes a prompt that instructs the second large language model to review the first chunk for a hallucination and a context that provides the first chunk and information dependent upon the topical information (Arunachalam; Fig. 5; [0068] … Next, LLM 110C may receive a second chunk and the answer generated from the first chunk to generate a second answer. Next, LLM 110C may receive a third chunk and the answer generated using the first and second chunks. The process may continue until LLM 110C uses all chunks in the subset of chunks or until LLM 110C determines that the content of the answer no longer changes.
[0071] … Next, LLM 110C may receive a second chunk and a summary generated from the first chunk to generate a summary. Next, LLM 110C may receive a third chunk and the summary generated using the first and second chunks. The process may continue until LLM 110C uses all chunks in the subset of chunks or until LLM 110C determines that the content of the summary is no longer being modified.); and
determining, by the second large language model, whether the first chunk of text includes at least one hallucination that includes information inconsistent with the topical information or the desired purpose of the first chunk (Arunachalam; [0045] Notably, in an embodiment in FIG. 2, a first instance of LLM 110 is finetuned on project data 202A to generate LLM 208A, a second instance of LLM 110 is finetuned on project data 202B to generate LLM 208B, and a third instance of LLM 110 is finetuned on project data 202C to generate LLM 208C. In this way, LLMs 208A, 208B, and 208C include respective project data 202A, 202B, and 202C, which reduces the likelihood of LLMs 208A-C hallucinating, or the likelihood of project data 202A-202C being intermingled with data from other projects when LLMs 208A-C generate a response to a question during the inference stage discussed in FIG. 3.
[0076] In some embodiments, generative AI system 108 may reduce or eliminate a number of AI hallucinations from the response. An AI hallucination may occur when one of LLMs 110 may create an answer or a summary that is not based on chunks 404. To reduce the likelihood of the hallucinations, the response displayed in session 502 of generative AI chatbot interface 116 may also include chunks 404 and/or original documents 204 and/or transcripts 206, or links to chunks 404 and/or original documents 204 and/or transcripts 206 from which LLMs 110 generated a response.)
Claim 16
Arunachalam also teaches the second large language model is different from the first large language model (Arunachalam; [0045] Notably, in an embodiment in FIG. 2, a first instance of LLM 110 is finetuned on project data 202A to generate LLM 208A, a second instance of LLM 110 is finetuned on project data 202B to generate LLM 208B, and a third instance of LLM 110 is finetuned on project data 202C to generate LLM 208C. In this way, LLMs 208A, 208B, and 208C include respective project data 202A, 202B, and 202C, which reduces the likelihood of LLMs 208A-C hallucinating, or the likelihood of project data 202A-202C being intermingled with data from other projects when LLMs 208A-C generate a response to a question during the inference stage discussed in FIG. 3.)
Claim 17
Arunachalam also teaches discarding the first chunk and repeating the process to generate a new first chunk (Arunachalam; [0076] In some embodiments, generative AI system 108 may reduce or eliminate a number of AI hallucinations from the response. An AI hallucination may occur when one of LLMs 110 may create an answer or a summary that is not based on chunks 404. To reduce the likelihood of the hallucinations, the response displayed in session 502 of generative AI chatbot interface 116 may also include chunks 404 and/or original documents 204 and/or transcripts 206, or links to chunks 404 and/or original documents 204 and/or transcripts 206 from which LLMs 110 generated a response.)
Claim 18
Arunachalam also teaches saving the first chunk in storage media (Chunks 404 are stored in memory)
Claim 19
Arunachalam also teaches initiating an alert that the first chunk includes at least one hallucination (Arunachalam; [0076] In some embodiments, generative AI system 108 may reduce or eliminate a number of AI hallucinations from the response. An AI hallucination may occur when one of LLMs 110 may create an answer or a summary that is not based on chunks 404…)
Claim 20
Arunachalam teaches before prompting the first large language model, assembling the first request including , the context, and the at least one example first chunk of text by a prompt module that at least partially includes the computer processor (Arunachalam; Fig. 5; [0066 – 0068] If information request router 504 determines that the information request is a question/answer request, information request router 504 passes the information request to vector retrieving module 506
One of LLM models, such as LLM 110C may receive one or more chunks in the subset of chunks to generate an answer to the question in the information request … One of LLM models, such as LLM 110C may receive one or more chunks in the subset of chunks to generate an answer to the question in the information request …; also see Fig. 7 & [0083 – 0091] a method for generating answer to question
[0071] If information request router 504 determines that the information request is a request for a summary, information request router 504 passes the information request to chunk retrieving module 510. Chunk retrieving module 510 may parse the metadata tags in dictionary 410A associated with project data 202A to determine tags that indicate that the corresponding chunk may include information that contributes to the summary …; also see Fig. 8 & [0092 – 0098]: a method for generating response to information request); and
providing the first request to the first large language model
One of LLM models, such as LLM 110C may receive one or more chunks in the subset of chunks to generate an answer to the question in the information request … One of LLM models, such as LLM 110C may receive one or more chunks in the subset of chunks to generate an answer to the question in the information request …; also see Fig. 7 & [0083 – 0091] a method for generating answer to question
[0071] If information request router 504 determines that the information request is a request for a summary, information request router 504 passes the information request to chunk retrieving module 510. Chunk retrieving module 510 may parse the metadata tags in dictionary 410A associated with project data 202A to determine tags that indicate that the corresponding chunk may include information that contributes to the summary …; also see Fig. 8 & [0092 – 0098]: a method for generating response to information request.)
Crouch teaches assembling the first request including the prompt;
providing the first request by the prompt module (Crouch; Fig. 5; [0075 – 0076] In step 506, the computing device may receive a prompt requiring passage search to be performed …
In step 508, the computing device may generate a search query from the received prompt for passage search. The computing device may normalize all received prompts into a search query composed of logical operators, search keywords, concept identifiers in a format (desired purpose) that is used to conduct passage (first chunk) searches by the QA system …, Also see [0079].) Motivation for incorporating Crouch into Arunachalam is the same as motivation in claim 1.
Claim 9 rejected under 35 U.S.C. 103 as being unpatentable over Arunachalam and Crouch as applied to claim 8 above, and further in view of Somaiya et al. (Pub. No. US 2024/0273306 A1; hereinafter Somaiya.)
Claim 9
Arunachalam teaches the multiple chunks corresponding headings (Arunachalam; [0055] Data processing engine 114 may also generate a metadata tags 406A, and associate each metadata tag in metadata tags 406A with each chunk in chunks 404A … The metadata tag may also include a chunk identifier …)
But Arunachalam and Crouch do not explicitly teach the multiple chunks are separate from one another within the document by corresponding headings.
However, Somaiya teaches the multiple chunks are separate from one another within the document by corresponding headings (Somaiya; [0280] In the example of FIG. 17, the example contributor-document interface display 1700 includes a display of a generative language model-generated document that includes contributions to specific segments of the document …
[0283] The segment 1706 includes a segment title 1720. In some implementations, segment title 1720 is machine-generated by a generative language model …
[0286] The segment 1708 includes a segment title 1734. In some implementations, segment title 1734 is machine-generated by a generative language model …
[0288] The segment 1710 includes a segment title 1740. In some implementations, segment title 1740 is machine-generated by a generative language model …)
Arunachalam, Crouch and Somaiya are in the same analogous art as they are in the same field of endeavor, generating documents. Therefore, it would have been obvious to a person of ordinary skill, in the art before the effective filing date of the claimed invention, to incorporate Somaiya into Arunachalam/Crouch to display chunks of texts along with their own title, as suggested by Crouch ([0283, 0286, & 0288]), to arrive at the invention.
Claim 19 rejected under 35 U.S.C. 103 as being unpatentable over Arunachalam and Crouch as applied to claim 15 above, and further in view of SOMECH et al. (Pub. No. US 2024/0419912 A1; hereinafter Somech.)
Claim 19
Arunachalam and Crouch do not teach initiating an alert that the first chunk includes at least one hallucination.
However, Somech teaches initiating an alert that the first chunk includes at least one hallucination (Somech; [0029] … the model output (e.g., generative text) may be supplemented with a hallucination score indicating that the likelihood of hallucination in the output is high, which gives a more accurate output result as opposed to a mere generative text output without any indication of hallucination.
[0106] FIG. 6 is a screenshot 600 of an example user interface for alerting users of a high likelihood of hallucination in a given output, according to some embodiments …)
Arunachalam, Crouch and Somech are in the same analogous art as they are in the same field of endeavor, generating documents. Therefore, it would have been obvious to a person of ordinary skill, in the art before the effective filing date of the claimed invention, to incorporate Somech into Arunachalam/Crouch to allow Arunachalam to alert user that texts may have hallucination based on hallucination score of the texts, as suggested by Somech ([0029 & 0106].)
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to CUONG V LUU whose telephone number is (571)270-1733. The examiner can normally be reached 6:30 AM - 3:00 PM.
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, Hyung S. Sough can be reached at (571) 272-6799. 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.
/CUONG V LUU/Examiner, Art Unit 2192
/S. Sough/SPE, Art Unit 2192
Prosecution Insights