DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Drawings
The drawings are objected to as failing to comply with 37 CFR 1.84(p)(5) because they include the following reference character(s) not mentioned in the description: 314 and 328. Corrected drawing sheets in compliance with 37 CFR 1.121(d), or amendment to the specification to add the reference character(s) in the description in compliance with 37 CFR 1.121(b) are required in reply to the Office action to avoid abandonment of the application. Any amended replacement drawing sheet should include all of the figures appearing on the immediate prior version of the sheet, even if only one figure is being amended. Each drawing sheet submitted after the filing date of an application must be labeled in the top margin as either “Replacement Sheet” or “New Sheet” pursuant to 37 CFR 1.121(d). If the changes are not accepted by the examiner, the applicant will be notified and informed of any required corrective action in the next Office action. The objection to the drawings will not be held in abeyance.
Claim Objections
Claim 1 is objected to because of the following informalities: Claim 1 recites “determining an amount non-overlapping content,” which is grammatically incorrect. Examiner suggests revising the claim to recite “determining an amount of non-overlapping content” (emphasis added). Appropriate correction is required.
Claim 8 is objected to because of the following informalities: Claim 8 recites “determining an amount non-overlapping content,” which is grammatically incorrect. Examiner suggests revising the claim to recite “determining an amount of non-overlapping content” (emphasis added). Appropriate correction is required.
Claim 15 is objected to because of the following informalities: Claim 15 recites “determine an amount non-overlapping content,” which is grammatically incorrect. Examiner suggests revising the claim to recite “determine an amount of non-overlapping content” (emphasis added). Appropriate correction is required.
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, 6-8, 13-15, and 20 are provisionally rejected on the ground of nonstatutory double patenting as being unpatentable over claims 1-3, 8-10, and 15-17 of copending Application No. 18/732,372 in view of Engi et al. (US Publication No. 2025/0371066).
Instant Application
U.S. Application No. 18/732,372
Claim
Limitation
Claim
Limitation
1
A computer-implemented method, executed on a computing device, comprising: generating a plurality of chunks for a plurality of text portions of a document;
3
The computer-implemented method of claim 1, further comprising: generating the plurality of chunks for a plurality of text portions of the target document;
generating a plurality of chunk embeddings from the plurality of chunks;
generating a plurality of chunk topics by extracting a topic for each respective chunk of the plurality of chunks; generating a weighting for the topic for each respective chunk of the plurality of chunks; and generating a plurality of weighted chunk topic embeddings by generating a weighted chunk topic embedding for each chunk.
processing a query using a generative artificial intelligence (AI) model;
1
A computer-implemented method, executed on a computing device, comprising: processing a query using a generative artificial intelligence (Al) model;
generating a query embedding from the query;
extracting a topic of the query; generating a weighting for the topic of the query by generating a weight for each term based on each term's frequency in a particular chunk from a plurality of chunks of a target document and each term's frequency across all of the plurality of chunks; generating a weighted query topic embedding for the topic of the query;
identifying a plurality of candidate chunks from the plurality of chunks based upon, at least in part, a similarity between the plurality of chunk embeddings and the query embedding;
identifying a candidate chunk from the plurality of chunks document by determining a similarity between the weighted query topic embedding and a plurality of chunk embeddings for the plurality of chunks;
2
The computer-implemented method of claim 1, wherein identifying a candidate chunk includes identifying a predefined number of most similar candidate chunks.
selecting a subset of the plurality of candidate chunks for inclusion in a prompt with the query.
1
generating a prompt using the query and the candidate chunk;
6
The computer-implemented method of claim 1, wherein selecting the subset of the plurality of candidate chunks for inclusion in the prompt includes: selecting a first candidate chunk with a highest similarity value compared to the query;
1
identifying a candidate chunk from the plurality of chunks document by determining a similarity between the weighted query topic embedding and a plurality of chunk embeddings for the plurality of chunks;
2
The computer-implemented method of claim 1, wherein identifying a candidate chunk includes identifying a predefined number of most similar candidate chunks.
7
The computer-implemented method of claim 1, further comprising: generating the prompt using the query and the subset of the plurality of candidate chunks; and
1
generating a prompt using the query and the candidate chunk;
providing the prompt to the generative AI model.
providing the prompt to the generative Al model.
The claims of U.S. Application No. 18/732,372 do not specifically disclose determining an amount non-overlapping content of each candidate chunk relative to each other candidate chunk; and selecting a subset of the plurality of candidate chunks for inclusion in a prompt based upon, at least in part, the amount of non-overlapping content of each candidate chunk. However, Engi teaches
determining an amount non-overlapping content of each candidate chunk relative to each other candidate chunk (see e.g., [0029] for in a step 313, the system applying cosine similarity, or a cosine similarity analysis, to the original prompt to identify a plurality of candidate chunks of documents, including one or more chunk nodes, that substantially match the original prompt, [0030] for after the plurality of candidate chunks is identified, the system analyzing one or more chunk nodes to identify chunk nodes with a relatively high centrality to the candidate chunks in a step 317, [0032] for similar information including, but not being limited to including, information that addresses the same general topic and/or information that may be described using substantially the same text, [0051] for method or step 317 of analyzing one or more candidate chunk nodes begins at a step 609 in which the system applies a graph traversal method to identify substantially all possible pathways from each candidate chunk to every other candidate chunk, [0052] for in a step 613, the system identifying a knowledge graph or subgraph of topics relative to the candidate chunks, using the knowledge graph, the system evaluating combinations of pathways or chunk paths using at least one metric to determine whether one or more common nodes exists between candidate chunks in a step 617, the combinations of pathways being evaluated to determine if there are nodes, e.g., nodes corresponding to ideas or topics, which are common between the candidate chunks, for example, common nodes traversed when evaluating paths between two candidate chunks being identified as idea nodes, and the metrics used to determine whether one or more common nodes exists may vary and may include, but are not limited to including, centrality, traversal, and/or connectivity. Identifying chunk nodes with a relatively high centrality to the candidate chunks includes determining common nodes existing between candidate chunks. The existence of a common node between two candidate chunks indicates that the two candidate chunks include information that addresses the same general topic and/or information that may be described using substantially the same text. The lack of existence of a common node between two candidate chunks indicates that the two candidate chunks contain information that address different topics and/or information that may be described using different text. Thus, an amount of non-overlapping content is determined for each candidate chunk relative to each other candidate chunk.); and
selecting a subset of the plurality of candidate chunks for inclusion in a prompt based upon, at least in part, the amount of non-overlapping content of each candidate chunk (see e.g., [0031] for a determination then made being in a step 321 as to whether one or more chunk nodes with relatively high centrality have been identified, if the determination is that there are essentially no chunk nodes with a relatively high centrality, the implication being that the original prompt provided sufficient information to allow a substantially relevant and/or accurate response to be provided, and that additional information not being needed from a user, and accordingly, process flow moving from step 321 to a step 325 in which the system provides a response to the user based on the context associated with the candidate chunks, [0032] for alternatively, if it is determined in step 321 that one or more chunk nodes with a relatively high centrality have been identified, the indication being that soliciting additional information from the user may enable a more relevant and/or accurate response to be provided to the user, and as such, in a step 329, the system utilizing the one or more chunk nodes with a relatively high centrality to identify paths or pathways to candidate chunks that contain similar information, [0033] for once pathways to candidate chunks that contain similar information are identified, the system generating at least one query or question in a step 333 configured to solicit clarification with respect to similar information, [0034] for in a step 337, the system presenting the query to the user, e.g., through a user interface such as a chatbot interface and the user providing a response, or a response prompt, to the query in a step 341, [0035] for after the system obtains the response prompt, the system analyzing candidate chunks based on the response prompt in a step 345, [0036] for the system dropping candidate chunks that are not consistent with or otherwise associated with the response prompt in a step 349, [0037] for from step 349, process flow moving to a step 353 in which the system provides remaining candidate chunks to an LLM arrangement, and [0038] for based on the remaining candidate chunks, the LLM arrangement effectively generating a response that the system provides to the user in a step 357. A subset of the candidate chunks is selected for prompting the LLM based upon the determination of common nodes between candidate chunks. Thus, the subset of the candidate chunks is selected for prompting the LLM based upon the amount of non-overlapping content.).
It would have been obvious to one of ordinary skill in the art prior to the effective filing date of the claimed invention to modify the method of the claims of U.S. Application No. 18/732,372 to determine an amount non-overlapping content of each candidate chunk relative to each other candidate chunk; and select a subset of the plurality of candidate chunks for inclusion in a prompt based upon, at least in part, the amount of non-overlapping content of each candidate chunk, as taught by Engi, for the benefit of ensuring the best context is used when querying an LLM (see e.g., Engi, [0004]).
As to claim 6, the limitations of parent claim 1 have been discussed above. The claims of U.S. Application No. 18/732,372 do not specifically disclose wherein selecting the subset of the plurality of candidate chunks for inclusion in the prompt includes: selecting an additional candidate chunk with a highest amount of non-overlapping content relative to remaining candidate chunks of the plurality of chunks. However, Engi teaches wherein selecting the subset of the plurality of candidate chunks for inclusion in the prompt includes:
selecting an additional candidate chunk [candidate chunk consistent with the response prompt] with a highest amount of non-overlapping content relative to remaining candidate chunks [candidate chunks that are not consistent with the response prompt] of the plurality of chunks (see e.g., [0032] for if it is determined in step 321 that one or more chunk nodes with a relatively high centrality have been identified, the indication being that soliciting additional information from the user may enable a more relevant and/or accurate response to be provided to the user, as such, in a step 329, the system utilizing the one or more chunk nodes with a relatively high centrality to identify paths or pathways to candidate chunks that contain similar information, and similar information including, but not being limited to including, information that addresses the same general topic and/or information that may be described using substantially the same text, [0033] for once pathways to candidate chunks that contain similar information are identified, the system generating at least one query or question in a step 333 configured to solicit clarification with respect to similar information, [0034] for in a step 337, the system presenting the query to the user, e.g., through a user interface such as a chatbot interface and the user providing a response, or a response prompt, to the query in a step 341, [0035] for after the system obtains the response prompt, the system analyzing candidate chunks based on the response prompt in a step 345, [0036] for the system dropping candidate chunks that are not consistent with or otherwise associated with the response prompt in a step 349, [0037] for from step 349, process flow moving to a step 353 in which the system provides remaining candidate chunks to an LLM arrangement, and [0038] for based on the remaining candidate chunks, the LLM arrangement effectively generating a response that the system provides to the user in a step 357. Candidate chunks consistent with the response prompt are selected, which includes a candidate chunk with a highest amount of non-overlapping content relative to candidate chunks that are not consistent with the response prompt. This candidate chunk is on a different path than candidate chunks that are not consistent with the response prompt, which means it contains dissimilar information from candidate chunks that are not consistent with the response prompt.).
It would have been obvious to one of ordinary skill in the art prior to the effective filing date of the claimed invention to modify the method of the claims of U.S. Application No. 18/732,372 wherein selecting the subset of the plurality of candidate chunks for inclusion in the prompt includes: selecting an additional candidate chunk with a highest amount of non-overlapping content relative to remaining candidate chunks of the plurality of chunks, as taught by Engi, for the benefit of ensuring the best context is used when querying an LLM (see e.g., Engi, [0004]).
This is a provisional nonstatutory double patenting rejection because the patentably indistinct claims have not in fact been patented.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
This application currently names joint inventors. In considering patentability of the claims the examiner presumes that the subject matter of the various claims was commonly owned as of the effective filing date of the claimed invention(s) absent any evidence to the contrary. Applicant is advised of the obligation under 37 CFR 1.56 to point out the inventor and effective filing dates of each claim that was not commonly owned as of the effective filing date of the later invention in order for the examiner to consider the applicability of 35 U.S.C. 102(b)(2)(C) for any potential 35 U.S.C. 102(a)(2) prior art against the later invention.
Claims 1, 2, 6-9, 13-16, and 20 are rejected under 35 U.S.C. 103 as being unpatentable over Mondlock et al. (US Patent No. 12,079,570) in view of Engi et al. (US Publication No. 2025/0371066).
As to claim 1, Mondlock teaches a computer-implemented method, executed on a computing device, comprising:
generating a plurality of chunks for a plurality of text portions of a document (see e.g., col. 15, lines 43-60 for splitting each selected document into a plurality of text chunks, col. 19, lines 20-25 for causing each document of the document set to be split into a plurality of text chunks, and col. 22, lines 60-67 for causing the plurality of documents to be split into a text chunks set);
generating a plurality of chunk embeddings from the plurality of chunks (see e.g., col. 15, lines 52-60 for generating an embedding for each text chunk and col. 23, lines 4-10 for causing text embeddings to be generated from the text chunks);
processing a query using a generative artificial intelligence (AI) model [LLM] (see e.g., col. 5, lines 14-31 for the LLM service 170 including an LLM model and an LLM being a type of artificial intelligence (AI) algorithm that uses deep learning techniques to perform a number of natural language processing (NLP) tasks, such as understanding, summarizing, generating, and/or predicting new content and col. 7, lines 52 for the intent classification module 126 including instructions for determining an intent of the query, the intent classification module 126 receiving the query or the query plus retrieved chat history and classifying the query into one or more of a plurality of pre-defined intents, and the intent classification module 126 (via the LLM interface module 132) submitting the user query and the pre-defined intents to the LLM service 170, e.g., using the text-davinci-003 model, and receiving intents as output);
generating a query embedding from the query (see e.g., col. 14, lines 41-46 for generating an embedding of the user query and col. 23, lines 19-25 for causing a query embedding to be generated from the user query);
identifying a plurality of candidate chunks [top relevant text chunks] from the plurality of chunks based upon, at least in part, a similarity between the plurality of chunk embeddings and the query embedding (see e.g., col. 15, line 61 – col. 16, line 3 for identifying the top relevant text chunks, the top relevant text chunks being identified by a semantic search, and the semantic search comprising performing a KNN search of the user query embedding and text chunk and/or data chunk embeddings. The KNN search identifies top relevant text chunks based on a similarity between the chunk embeddings and the query embedding.); and
selecting a subset of the plurality of candidate chunks for inclusion in a prompt with the query (see e.g., col. 16, lines 13-37 for the LLM interface module 132 submitting a prompt to the LLM service 170 comprising the user query and the top relevant text chunks and in some aspects, a plurality of prompts being submitted to the LLM service 170 comprising one or more of the text chunks).
Mondlock does not specifically disclose determining an amount non-overlapping content of each candidate chunk relative to each other candidate chunk; and selecting a subset of the plurality of candidate chunks for inclusion in a prompt based upon, at least in part, the amount of non-overlapping content of each candidate chunk. However, Engi teaches
determining an amount non-overlapping content of each candidate chunk relative to each other candidate chunk (see e.g., [0029] for in a step 313, the system applying cosine similarity, or a cosine similarity analysis, to the original prompt to identify a plurality of candidate chunks of documents, including one or more chunk nodes, that substantially match the original prompt, [0030] for after the plurality of candidate chunks is identified, the system analyzing one or more chunk nodes to identify chunk nodes with a relatively high centrality to the candidate chunks in a step 317, [0032] for similar information including, but not being limited to including, information that addresses the same general topic and/or information that may be described using substantially the same text, [0051] for method or step 317 of analyzing one or more candidate chunk nodes begins at a step 609 in which the system applies a graph traversal method to identify substantially all possible pathways from each candidate chunk to every other candidate chunk, [0052] for in a step 613, the system identifying a knowledge graph or subgraph of topics relative to the candidate chunks, using the knowledge graph, the system evaluating combinations of pathways or chunk paths using at least one metric to determine whether one or more common nodes exists between candidate chunks in a step 617, the combinations of pathways being evaluated to determine if there are nodes, e.g., nodes corresponding to ideas or topics, which are common between the candidate chunks, for example, common nodes traversed when evaluating paths between two candidate chunks being identified as idea nodes, and the metrics used to determine whether one or more common nodes exists may vary and may include, but are not limited to including, centrality, traversal, and/or connectivity. Identifying chunk nodes with a relatively high centrality to the candidate chunks includes determining common nodes existing between candidate chunks. The existence of a common node between two candidate chunks indicates that the two candidate chunks include information that addresses the same general topic and/or information that may be described using substantially the same text. The lack of existence of a common node between two candidate chunks indicates that the two candidate chunks contain information that address different topics and/or information that may be described using different text. Thus, an amount of non-overlapping content is determined for each candidate chunk relative to each other candidate chunk.); and
selecting a subset of the plurality of candidate chunks for inclusion in a prompt based upon, at least in part, the amount of non-overlapping content of each candidate chunk (see e.g., [0031] for a determination then made being in a step 321 as to whether one or more chunk nodes with relatively high centrality have been identified, if the determination is that there are essentially no chunk nodes with a relatively high centrality, the implication being that the original prompt provided sufficient information to allow a substantially relevant and/or accurate response to be provided, and that additional information not being needed from a user, and accordingly, process flow moving from step 321 to a step 325 in which the system provides a response to the user based on the context associated with the candidate chunks, [0032] for alternatively, if it is determined in step 321 that one or more chunk nodes with a relatively high centrality have been identified, the indication being that soliciting additional information from the user may enable a more relevant and/or accurate response to be provided to the user, and as such, in a step 329, the system utilizing the one or more chunk nodes with a relatively high centrality to identify paths or pathways to candidate chunks that contain similar information, [0033] for once pathways to candidate chunks that contain similar information are identified, the system generating at least one query or question in a step 333 configured to solicit clarification with respect to similar information, [0034] for in a step 337, the system presenting the query to the user, e.g., through a user interface such as a chatbot interface and the user providing a response, or a response prompt, to the query in a step 341, [0035] for after the system obtains the response prompt, the system analyzing candidate chunks based on the response prompt in a step 345, [0036] for the system dropping candidate chunks that are not consistent with or otherwise associated with the response prompt in a step 349, [0037] for from step 349, process flow moving to a step 353 in which the system provides remaining candidate chunks to an LLM arrangement, and [0038] for based on the remaining candidate chunks, the LLM arrangement effectively generating a response that the system provides to the user in a step 357. A subset of the candidate chunks is selected for prompting the LLM based upon the determination of common nodes between candidate chunks. Thus, the subset of the candidate chunks is selected for prompting the LLM based upon the amount of non-overlapping content.).
It would have been obvious to one of ordinary skill in the art prior to the effective filing date of the claimed invention to modify the RAG pipeline of Mondlock to determine an amount non-overlapping content of each candidate chunk relative to each other candidate chunk; and select a subset of the plurality of candidate chunks for inclusion in a prompt based upon, at least in part, the amount of non-overlapping content of each candidate chunk, as taught by Engi, for the benefit of ensuring the best context is used when querying an LLM (see e.g., Engi, [0004]).
As to claim 2, the limitations of parent claim 1 have been discussed above. Mondlock teaches
wherein processing the query includes processing the query during Retrieval Augmented Generation (RAG) using the generative AI model (see e.g., col. 7, lines 52 for the intent classification module 126 (via the LLM interface module 132) submitting the user query and the pre-defined intents to the LLM service 170, e.g., using the text-davinci-003 model, and receiving intents as output and col. 11, lines 59-67 for if the determination is to use RAG then the RAG pipeline 300 including at block 332 detecting the intent of the user query and the intent detection being performed by the intent classification module. The LLM processes the query during RAG.).
As to claim 6, the limitations of parent claim 1 have been discussed above. Mondlock teaches wherein selecting the subset of the plurality of candidate chunks for inclusion in the prompt includes:
selecting a first candidate chunk [top relevant text chunk] with a highest similarity value compared to the query (see e.g., col. 15, line 61 – col. 16, line 3 for identifying the top relevant text chunks, the top relevant text chunks being identified by a semantic search, and the semantic search comprising performing a KNN search of the user query embedding and text chunk and/or data chunk embeddings. The KNN search identifies top relevant text chunks based on a similarity between the chunk embeddings and the query embedding.).
Mondlock does not specifically disclose wherein selecting the subset of the plurality of candidate chunks for inclusion in the prompt includes: selecting an additional candidate chunk with a highest amount of non-overlapping content relative to remaining candidate chunks of the plurality of chunks. However, Engi teaches wherein selecting the subset of the plurality of candidate chunks for inclusion in the prompt includes:
selecting an additional candidate chunk [candidate chunk consistent with the response prompt] with a highest amount of non-overlapping content relative to remaining candidate chunks [candidate chunks that are not consistent with the response prompt] of the plurality of chunks (see e.g., [0032] for if it is determined in step 321 that one or more chunk nodes with a relatively high centrality have been identified, the indication being that soliciting additional information from the user may enable a more relevant and/or accurate response to be provided to the user, as such, in a step 329, the system utilizing the one or more chunk nodes with a relatively high centrality to identify paths or pathways to candidate chunks that contain similar information, and similar information including, but not being limited to including, information that addresses the same general topic and/or information that may be described using substantially the same text, [0033] for once pathways to candidate chunks that contain similar information are identified, the system generating at least one query or question in a step 333 configured to solicit clarification with respect to similar information, [0034] for in a step 337, the system presenting the query to the user, e.g., through a user interface such as a chatbot interface and the user providing a response, or a response prompt, to the query in a step 341, [0035] for after the system obtains the response prompt, the system analyzing candidate chunks based on the response prompt in a step 345, [0036] for the system dropping candidate chunks that are not consistent with or otherwise associated with the response prompt in a step 349, [0037] for from step 349, process flow moving to a step 353 in which the system provides remaining candidate chunks to an LLM arrangement, and [0038] for based on the remaining candidate chunks, the LLM arrangement effectively generating a response that the system provides to the user in a step 357. Candidate chunks consistent with the response prompt are selected, which includes a candidate chunk with a highest amount of non-overlapping content relative to candidate chunks that are not consistent with the response prompt. This candidate chunk is on a different path than candidate chunks that are not consistent with the response prompt, which means it contains dissimilar information from candidate chunks that are not consistent with the response prompt.).
It would have been obvious to one of ordinary skill in the art prior to the effective filing date of the claimed invention to modify the RAG pipeline of Mondlock wherein selecting the subset of the plurality of candidate chunks for inclusion in the prompt includes: selecting an additional candidate chunk with a highest amount of non-overlapping content relative to remaining candidate chunks of the plurality of chunks, as taught by Engi, for the benefit of ensuring the best context is used when querying an LLM (see e.g., Engi, [0004]).
As to claim 7, the limitations of parent claim 1 have been discussed above. Mondlock teaches
generating the prompt using the query and the subset of the plurality of candidate chunks (see e.g., col. 16, lines 13-37 for the LLM interface module 132 submitting a prompt to the LLM service 170 comprising the user query and the top relevant text chunks and in some aspects, a plurality of prompts being submitted to the LLM service 170 comprising one or more of the text chunks); and
providing the prompt to the generative AI model (see e.g., col. 16, lines 13-37 for the LLM interface module 132 submitting a prompt to the LLM service 170 comprising the user query and the top relevant text chunks and in some aspects, a plurality of prompts being submitted to the LLM service 170 comprising one or more of the text chunks).
As to claim 8, Mondlock teaches a computer product residing on a non-transitory computer readable medium having a plurality of instructions stored thereon which, when executed by a processor, cause the processor to perform operations comprising:
generating a plurality of chunks for a plurality of text portions of a document (see e.g., col. 15, lines 43-60 for splitting each selected document into a plurality of text chunks, col. 19, lines 20-25 for causing each document of the document set to be split into a plurality of text chunks, and col. 22, lines 60-67 for causing the plurality of documents to be split into a text chunks set);
generating a plurality of chunk embeddings from the plurality of chunks (see e.g., col. 15, lines 52-60 for generating an embedding for each text chunk and col. 23, lines 4-10 for causing text embeddings to be generated from the text chunks);
processing a query using a generative artificial intelligence (AI) model [LLM] (see e.g., col. 5, lines 14-31 for the LLM service 170 including an LLM model and an LLM being a type of artificial intelligence (AI) algorithm that uses deep learning techniques to perform a number of natural language processing (NLP) tasks, such as understanding, summarizing, generating, and/or predicting new content and col. 7, lines 52 for the intent classification module 126 including instructions for determining an intent of the query, the intent classification module 126 receiving the query or the query plus retrieved chat history and classifying the query into one or more of a plurality of pre-defined intents, and the intent classification module 126 (via the LLM interface module 132) submitting the user query and the pre-defined intents to the LLM service 170, e.g., using the text-davinci-003 model, and receiving intents as output);
generating a query embedding from the query (see e.g., col. 14, lines 41-46 for generating an embedding of the user query and col. 23, lines 19-25 for causing a query embedding to be generated from the user query);
identifying a plurality of candidate chunks [top relevant text chunks] from the plurality of chunks based upon, at least in part, a similarity between the plurality of chunk embeddings and the query embedding (see e.g., col. 15, line 61 – col. 16, line 3 for identifying the top relevant text chunks, the top relevant text chunks being identified by a semantic search, and the semantic search comprising performing a KNN search of the user query embedding and text chunk and/or data chunk embeddings. The KNN search identifies top relevant text chunks based on a similarity between the chunk embeddings and the query embedding.); and
selecting a subset of the plurality of candidate chunks for inclusion in a prompt with the query (see e.g., col. 16, lines 13-37 for the LLM interface module 132 submitting a prompt to the LLM service 170 comprising the user query and the top relevant text chunks and in some aspects, a plurality of prompts being submitted to the LLM service 170 comprising one or more of the text chunks).
Mondlock does not specifically disclose determining an amount non-overlapping content of each candidate chunk relative to each other candidate chunk; and selecting a subset of the plurality of candidate chunks for inclusion in a prompt based upon, at least in part, the amount of non-overlapping content of each candidate chunk. However, Engi teaches
determining an amount non-overlapping content of each candidate chunk relative to each other candidate chunk (see e.g., [0029] for in a step 313, the system applying cosine similarity, or a cosine similarity analysis, to the original prompt to identify a plurality of candidate chunks of documents, including one or more chunk nodes, that substantially match the original prompt, [0030] for after the plurality of candidate chunks is identified, the system analyzing one or more chunk nodes to identify chunk nodes with a relatively high centrality to the candidate chunks in a step 317, [0032] for similar information including, but not being limited to including, information that addresses the same general topic and/or information that may be described using substantially the same text, [0051] for method or step 317 of analyzing one or more candidate chunk nodes begins at a step 609 in which the system applies a graph traversal method to identify substantially all possible pathways from each candidate chunk to every other candidate chunk, [0052] for in a step 613, the system identifying a knowledge graph or subgraph of topics relative to the candidate chunks, using the knowledge graph, the system evaluating combinations of pathways or chunk paths using at least one metric to determine whether one or more common nodes exists between candidate chunks in a step 617, the combinations of pathways being evaluated to determine if there are nodes, e.g., nodes corresponding to ideas or topics, which are common between the candidate chunks, for example, common nodes traversed when evaluating paths between two candidate chunks being identified as idea nodes, and the metrics used to determine whether one or more common nodes exists may vary and may include, but are not limited to including, centrality, traversal, and/or connectivity. Identifying chunk nodes with a relatively high centrality to the candidate chunks includes determining common nodes existing between candidate chunks. The existence of a common node between two candidate chunks indicates that the two candidate chunks include information that addresses the same general topic and/or information that may be described using substantially the same text. The lack of existence of a common node between two candidate chunks indicates that the two candidate chunks contain information that address different topics and/or information that may be described using different text. Thus, an amount of non-overlapping content is determined for each candidate chunk relative to each other candidate chunk.); and
selecting a subset of the plurality of candidate chunks for inclusion in a prompt based upon, at least in part, the amount of non-overlapping content of each candidate chunk (see e.g., [0031] for a determination then made being in a step 321 as to whether one or more chunk nodes with relatively high centrality have been identified, if the determination is that there are essentially no chunk nodes with a relatively high centrality, the implication being that the original prompt provided sufficient information to allow a substantially relevant and/or accurate response to be provided, and that additional information not being needed from a user, and accordingly, process flow moving from step 321 to a step 325 in which the system provides a response to the user based on the context associated with the candidate chunks, [0032] for alternatively, if it is determined in step 321 that one or more chunk nodes with a relatively high centrality have been identified, the indication being that soliciting additional information from the user may enable a more relevant and/or accurate response to be provided to the user, and as such, in a step 329, the system utilizing the one or more chunk nodes with a relatively high centrality to identify paths or pathways to candidate chunks that contain similar information, [0033] for once pathways to candidate chunks that contain similar information are identified, the system generating at least one query or question in a step 333 configured to solicit clarification with respect to similar information, [0034] for in a step 337, the system presenting the query to the user, e.g., through a user interface such as a chatbot interface and the user providing a response, or a response prompt, to the query in a step 341, [0035] for after the system obtains the response prompt, the system analyzing candidate chunks based on the response prompt in a step 345, [0036] for the system dropping candidate chunks that are not consistent with or otherwise associated with the response prompt in a step 349, [0037] for from step 349, process flow moving to a step 353 in which the system provides remaining candidate chunks to an LLM arrangement, and [0038] for based on the remaining candidate chunks, the LLM arrangement effectively generating a response that the system provides to the user in a step 357. A subset of the candidate chunks is selected for prompting the LLM based upon the determination of common nodes between candidate chunks. Thus, the subset of the candidate chunks is selected for prompting the LLM based upon the amount of non-overlapping content.).
It would have been obvious to one of ordinary skill in the art prior to the effective filing date of the claimed invention to modify the RAG pipeline of Mondlock to determine an amount non-overlapping content of each candidate chunk relative to each other candidate chunk; and select a subset of the plurality of candidate chunks for inclusion in a prompt based upon, at least in part, the amount of non-overlapping content of each candidate chunk, as taught by Engi, for the benefit of ensuring the best context is used when querying an LLM (see e.g., Engi, [0004]).
As to claim 9, the limitations of parent claim 8 have been discussed above. Mondlock teaches
wherein processing the query includes processing the query during Retrieval Augmented Generation (RAG) using the generative AI model (see e.g., col. 7, lines 52 for the intent classification module 126 (via the LLM interface module 132) submitting the user query and the pre-defined intents to the LLM service 170, e.g., using the text-davinci-003 model, and receiving intents as output and col. 11, lines 59-67 for if the determination is to use RAG then the RAG pipeline 300 including at block 332 detecting the intent of the user query and the intent detection being performed by the intent classification module. The LLM processes the query during RAG.).
As to claim 13, the limitations of parent claim 8 have been discussed above. Mondlock teaches wherein selecting the subset of the plurality of candidate chunks for inclusion in the prompt includes:
selecting a first candidate chunk [top relevant text chunk] with a highest similarity value compared to the query (see e.g., col. 15, line 61 – col. 16, line 3 for identifying the top relevant text chunks, the top relevant text chunks being identified by a semantic search, and the semantic search comprising performing a KNN search of the user query embedding and text chunk and/or data chunk embeddings. The KNN search identifies top relevant text chunks based on a similarity between the chunk embeddings and the query embedding.).
Mondlock does not specifically disclose wherein selecting the subset of the plurality of candidate chunks for inclusion in the prompt includes: selecting an additional candidate chunk with a highest amount of non-overlapping content relative to remaining candidate chunks of the plurality of chunks. However, Engi teaches wherein selecting the subset of the plurality of candidate chunks for inclusion in the prompt includes:
selecting an additional candidate chunk [candidate chunk consistent with the response prompt] with a highest amount of non-overlapping content relative to remaining candidate chunks [candidate chunks that are not consistent with the response prompt] of the plurality of chunks (see e.g., [0032] for if it is determined in step 321 that one or more chunk nodes with a relatively high centrality have been identified, the indication being that soliciting additional information from the user may enable a more relevant and/or accurate response to be provided to the user, as such, in a step 329, the system utilizing the one or more chunk nodes with a relatively high centrality to identify paths or pathways to candidate chunks that contain similar information, and similar information including, but not being limited to including, information that addresses the same general topic and/or information that may be described using substantially the same text, [0033] for once pathways to candidate chunks that contain similar information are identified, the system generating at least one query or question in a step 333 configured to solicit clarification with respect to similar information, [0034] for in a step 337, the system presenting the query to the user, e.g., through a user interface such as a chatbot interface and the user providing a response, or a response prompt, to the query in a step 341, [0035] for after the system obtains the response prompt, the system analyzing candidate chunks based on the response prompt in a step 345, [0036] for the system dropping candidate chunks that are not consistent with or otherwise associated with the response prompt in a step 349, [0037] for from step 349, process flow moving to a step 353 in which the system provides remaining candidate chunks to an LLM arrangement, and [0038] for based on the remaining candidate chunks, the LLM arrangement effectively generating a response that the system provides to the user in a step 357. Candidate chunks consistent with the response prompt are selected, which includes a candidate chunk with a highest amount of non-overlapping content relative to candidate chunks that are not consistent with the response prompt. This candidate chunk is on a different path than candidate chunks that are not consistent with the response prompt, which means it contains dissimilar information from candidate chunks that are not consistent with the response prompt.).
It would have been obvious to one of ordinary skill in the art prior to the effective filing date of the claimed invention to modify the RAG pipeline of Mondlock wherein selecting the subset of the plurality of candidate chunks for inclusion in the prompt includes: selecting an additional candidate chunk with a highest amount of non-overlapping content relative to remaining candidate chunks of the plurality of chunks, as taught by Engi, for the benefit of ensuring the best context is used when querying an LLM (see e.g., Engi, [0004]).
As to claim 14, the limitations of parent claims 8 and 13 have been discussed above. Mondlock teaches wherein the operations further comprise:
generating the prompt using the query and the subset of the plurality of candidate chunks (see e.g., col. 16, lines 13-37 for the LLM interface module 132 submitting a prompt to the LLM service 170 comprising the user query and the top relevant text chunks and in some aspects, a plurality of prompts being submitted to the LLM service 170 comprising one or more of the text chunks); and
providing the prompt to the generative AI model (see e.g., col. 16, lines 13-37 for the LLM interface module 132 submitting a prompt to the LLM service 170 comprising the user query and the top relevant text chunks and in some aspects, a plurality of prompts being submitted to the LLM service 170 comprising one or more of the text chunks).
As to claim 15, Mondlock teaches a computing system comprising:
a memory (see e.g., [0024] for the one or more processors 112 being configured to execute software instructions stored in a memory);
a processor (see e.g., [0024] for the one or more processors 112 being configured to execute software instructions stored in a memory) configured to generate a plurality of chunks for a plurality of text portions of a document (see e.g., col. 15, lines 43-60 for splitting each selected document into a plurality of text chunks, col. 19, lines 20-25 for causing each document of the document set to be split into a plurality of text chunks, and col. 22, lines 60-67 for causing the plurality of documents to be split into a text chunks set); to generate a plurality of chunk embeddings from the plurality of chunks (see e.g., col. 15, lines 52-60 for generating an embedding for each text chunk and col. 23, lines 4-10 for causing text embeddings to be generated from the text chunks); to process a query using a generative artificial intelligence (AI) model [LLM] (see e.g., col. 5, lines 14-31 for the LLM service 170 including an LLM model and an LLM being a type of artificial intelligence (AI) algorithm that uses deep learning techniques to perform a number of natural language processing (NLP) tasks, such as understanding, summarizing, generating, and/or predicting new content and col. 7, lines 52 for the intent classification module 126 including instructions for determining an intent of the query, the intent classification module 126 receiving the query or the query plus retrieved chat history and classifying the query into one or more of a plurality of pre-defined intents, and the intent classification module 126 (via the LLM interface module 132) submitting the user query and the pre-defined intents to the LLM service 170, e.g., using the text-davinci-003 model, and receiving intents as output); to generate a query embedding from the query (see e.g., col. 14, lines 41-46 for generating an embedding of the user query and col. 23, lines 19-25 for causing a query embedding to be generated from the user query); to identify a plurality of candidate chunks [top relevant text chunks] from the plurality of chunks based upon, at least in part, a similarity between the plurality of chunk embeddings and the query embedding (see e.g., col. 15, line 61 – col. 16, line 3 for identifying the top relevant text chunks, the top relevant text chunks being identified by a semantic search, and the semantic search comprising performing a KNN search of the user query embedding and text chunk and/or data chunk embeddings. The KNN search identifies top relevant text chunks based on a similarity between the chunk embeddings and the query embedding.); and to select a subset of the plurality of candidate chunks for inclusion in a prompt with the query (see e.g., col. 16, lines 13-37 for the LLM interface module 132 submitting a prompt to the LLM service 170 comprising the user query and the top relevant text chunks and in some aspects, a plurality of prompts being submitted to the LLM service 170 comprising one or more of the text chunks).
Mondlock does not specifically disclose to determine an amount non-overlapping content of each candidate chunk relative to each other candidate chunk; and to select a subset of the plurality of candidate chunks for inclusion in a prompt based upon, at least in part, the amount of non-overlapping content of each candidate chunk. However, Engi teaches
to determine an amount non-overlapping content of each candidate chunk relative to each other candidate chunk (see e.g., [0029] for in a step 313, the system applying cosine similarity, or a cosine similarity analysis, to the original prompt to identify a plurality of candidate chunks of documents, including one or more chunk nodes, that substantially match the original prompt, [0030] for after the plurality of candidate chunks is identified, the system analyzing one or more chunk nodes to identify chunk nodes with a relatively high centrality to the candidate chunks in a step 317, [0032] for similar information including, but not being limited to including, information that addresses the same general topic and/or information that may be described using substantially the same text, [0051] for method or step 317 of analyzing one or more candidate chunk nodes begins at a step 609 in which the system applies a graph traversal method to identify substantially all possible pathways from each candidate chunk to every other candidate chunk, [0052] for in a step 613, the system identifying a knowledge graph or subgraph of topics relative to the candidate chunks, using the knowledge graph, the system evaluating combinations of pathways or chunk paths using at least one metric to determine whether one or more common nodes exists between candidate chunks in a step 617, the combinations of pathways being evaluated to determine if there are nodes, e.g., nodes corresponding to ideas or topics, which are common between the candidate chunks, for example, common nodes traversed when evaluating paths between two candidate chunks being identified as idea nodes, and the metrics used to determine whether one or more common nodes exists may vary and may include, but are not limited to including, centrality, traversal, and/or connectivity. Identifying chunk nodes with a relatively high centrality to the candidate chunks includes determining common nodes existing between candidate chunks. The existence of a common node between two candidate chunks indicates that the two candidate chunks include information that addresses the same general topic and/or information that may be described using substantially the same text. The lack of existence of a common node between two candidate chunks indicates that the two candidate chunks contain information that address different topics and/or information that may be described using different text. Thus, an amount of non-overlapping content is determined for each candidate chunk relative to each other candidate chunk.); and
to select a subset of the plurality of candidate chunks for inclusion in a prompt based upon, at least in part, the amount of non-overlapping content of each candidate chunk (see e.g., [0031] for a determination then made being in a step 321 as to whether one or more chunk nodes with relatively high centrality have been identified, if the determination is that there are essentially no chunk nodes with a relatively high centrality, the implication being that the original prompt provided sufficient information to allow a substantially relevant and/or accurate response to be provided, and that additional information not being needed from a user, and accordingly, process flow moving from step 321 to a step 325 in which the system provides a response to the user based on the context associated with the candidate chunks, [0032] for alternatively, if it is determined in step 321 that one or more chunk nodes with a relatively high centrality have been identified, the indication being that soliciting additional information from the user may enable a more relevant and/or accurate response to be provided to the user, and as such, in a step 329, the system utilizing the one or more chunk nodes with a relatively high centrality to identify paths or pathways to candidate chunks that contain similar information, [0033] for once pathways to candidate chunks that contain similar information are identified, the system generating at least one query or question in a step 333 configured to solicit clarification with respect to similar information, [0034] for in a step 337, the system presenting the query to the user, e.g., through a user interface such as a chatbot interface and the user providing a response, or a response prompt, to the query in a step 341, [0035] for after the system obtains the response prompt, the system analyzing candidate chunks based on the response prompt in a step 345, [0036] for the system dropping candidate chunks that are not consistent with or otherwise associated with the response prompt in a step 349, [0037] for from step 349, process flow moving to a step 353 in which the system provides remaining candidate chunks to an LLM arrangement, and [0038] for based on the remaining candidate chunks, the LLM arrangement effectively generating a response that the system provides to the user in a step 357. A subset of the candidate chunks is selected for prompting the LLM based upon the determination of common nodes between candidate chunks. Thus, the subset of the candidate chunks is selected for prompting the LLM based upon the amount of non-overlapping content.).
It would have been obvious to one of ordinary skill in the art prior to the effective filing date of the claimed invention to modify the RAG pipeline of Mondlock to determine an amount non-overlapping content of each candidate chunk relative to each other candidate chunk; and select a subset of the plurality of candidate chunks for inclusion in a prompt based upon, at least in part, the amount of non-overlapping content of each candidate chunk, as taught by Engi, for the benefit of ensuring the best context is used when querying an LLM (see e.g., Engi, [0004]).
As to claim 16, the limitations of parent claim 15 have been discussed above. Mondlock teaches
wherein processing the query includes processing the query during Retrieval Augmented Generation (RAG) using the generative AI model (see e.g., col. 7, lines 52 for the intent classification module 126 (via the LLM interface module 132) submitting the user query and the pre-defined intents to the LLM service 170, e.g., using the text-davinci-003 model, and receiving intents as output and col. 11, lines 59-67 for if the determination is to use RAG then the RAG pipeline 300 including at block 332 detecting the intent of the user query and the intent detection being performed by the intent classification module. The LLM processes the query during RAG.).
As to claim 20, the limitations of parent claim 15 have been discussed above. Mondlock teaches wherein selecting the subset of the plurality of candidate chunks for inclusion in the prompt includes:
selecting a first candidate chunk [top relevant text chunk] with a highest similarity value compared to the query (see e.g., col. 15, line 61 – col. 16, line 3 for identifying the top relevant text chunks, the top relevant text chunks being identified by a semantic search, and the semantic search comprising performing a KNN search of the user query embedding and text chunk and/or data chunk embeddings. The KNN search identifies top relevant text chunks based on a similarity between the chunk embeddings and the query embedding.).
Mondlock does not specifically disclose wherein selecting the subset of the plurality of candidate chunks for inclusion in the prompt includes: selecting an additional candidate chunk with a highest amount of non-overlapping content relative to remaining candidate chunks of the plurality of chunks. However, Engi teaches wherein selecting the subset of the plurality of candidate chunks for inclusion in the prompt includes:
selecting an additional candidate chunk [candidate chunk consistent with the response prompt] with a highest amount of non-overlapping content relative to remaining candidate chunks [candidate chunks that are not consistent with the response prompt] of the plurality of chunks (see e.g., [0032] for if it is determined in step 321 that one or more chunk nodes with a relatively high centrality have been identified, the indication being that soliciting additional information from the user may enable a more relevant and/or accurate response to be provided to the user, as such, in a step 329, the system utilizing the one or more chunk nodes with a relatively high centrality to identify paths or pathways to candidate chunks that contain similar information, and similar information including, but not being limited to including, information that addresses the same general topic and/or information that may be described using substantially the same text, [0033] for once pathways to candidate chunks that contain similar information are identified, the system generating at least one query or question in a step 333 configured to solicit clarification with respect to similar information, [0034] for in a step 337, the system presenting the query to the user, e.g., through a user interface such as a chatbot interface and the user providing a response, or a response prompt, to the query in a step 341, [0035] for after the system obtains the response prompt, the system analyzing candidate chunks based on the response prompt in a step 345, [0036] for the system dropping candidate chunks that are not consistent with or otherwise associated with the response prompt in a step 349, [0037] for from step 349, process flow moving to a step 353 in which the system provides remaining candidate chunks to an LLM arrangement, and [0038] for based on the remaining candidate chunks, the LLM arrangement effectively generating a response that the system provides to the user in a step 357. Candidate chunks consistent with the response prompt are selected, which includes a candidate chunk with a highest amount of non-overlapping content relative to candidate chunks that are not consistent with the response prompt. This candidate chunk is on a different path than candidate chunks that are not consistent with the response prompt, which means it contains dissimilar information from candidate chunks that are not consistent with the response prompt.).
It would have been obvious to one of ordinary skill in the art prior to the effective filing date of the claimed invention to modify the RAG pipeline of Mondlock wherein selecting the subset of the plurality of candidate chunks for inclusion in the prompt includes: selecting an additional candidate chunk with a highest amount of non-overlapping content relative to remaining candidate chunks of the plurality of chunks, as taught by Engi, for the benefit of ensuring the best context is used when querying an LLM (see e.g., Engi, [0004]).
Claims 3, 10, and 17 are rejected under 35 U.S.C. 103 as being unpatentable over Mondlock et al. (US Patent No. 12,079,570) in view of Engi et al. (US Publication No. 2025/0371066) as applied to claims 1, 2, 6-9, 13-16, and 20 above, and further in view of Sivulka (US Patent No. 12,271,707).
As to claim 3, the limitations of parent claim 1 have been discussed above. Mondlock in view of Engi does not specifically disclose wherein determining the amount of non-overlapping content of each candidate chunk includes determining a cosine-similarity between each candidate chunk relative to each other candidate chunk. However, Sivulka teaches
wherein determining the amount of non-overlapping content of each candidate chunk includes determining a cosine-similarity between each candidate chunk relative to each other candidate chunk (see e.g., col. 12, lines 1-11 for the data analytics system iteratively comparing section embeddings through the sequence of sections in a source and designating the end of a chunk when the difference between the embeddings of two sections exceeds a threshold value and for example, the data analytics system starting a new chunk when the distance or cosine similarity between successive chunk embeddings exceeds some threshold value).
It would have been obvious to one of ordinary skill in the art prior to the effective filing date of the claimed invention to modify the RAG pipeline of Mondlock in view of Engi wherein determining the amount of non-overlapping content of each candidate chunk includes determining a cosine-similarity between each candidate chunk relative to each other candidate chunk, as taught by Sivulka, because cosine similarity is one of a finite number of identified, predictable similarity metrics with a reasonable expectation of success, and thus would have been obvious to try.
As to claim 10, the limitations of parent claim 8 have been discussed above. Mondlock in view of Engi does not specifically disclose wherein determining the amount of non-overlapping content of each candidate chunk includes determining a cosine-similarity between each candidate chunk relative to each other candidate chunk. However, Sivulka teaches
wherein determining the amount of non-overlapping content of each candidate chunk includes determining a cosine-similarity between each candidate chunk relative to each other candidate chunk (see e.g., col. 12, lines 1-11 for the data analytics system iteratively comparing section embeddings through the sequence of sections in a source and designating the end of a chunk when the difference between the embeddings of two sections exceeds a threshold value and for example, the data analytics system starting a new chunk when the distance or cosine similarity between successive chunk embeddings exceeds some threshold value).
It would have been obvious to one of ordinary skill in the art prior to the effective filing date of the claimed invention to modify the RAG pipeline of Mondlock in view of Engi wherein determining the amount of non-overlapping content of each candidate chunk includes determining a cosine-similarity between each candidate chunk relative to each other candidate chunk, as taught by Sivulka, because cosine similarity is one of a finite number of identified, predictable similarity metrics with a reasonable expectation of success, and thus would have been obvious to try.
As to claim 17, the limitations of parent claim 15 have been discussed above. Mondlock in view of Engi does not specifically disclose wherein determining the amount of non-overlapping content of each candidate chunk includes determining a cosine-similarity between each candidate chunk relative to each other candidate chunk. However, Sivulka teaches
wherein determining the amount of non-overlapping content of each candidate chunk includes determining a cosine-similarity between each candidate chunk relative to each other candidate chunk (see e.g., col. 12, lines 1-11 for the data analytics system iteratively comparing section embeddings through the sequence of sections in a source and designating the end of a chunk when the difference between the embeddings of two sections exceeds a threshold value and for example, the data analytics system starting a new chunk when the distance or cosine similarity between successive chunk embeddings exceeds some threshold value).
It would have been obvious to one of ordinary skill in the art prior to the effective filing date of the claimed invention to modify the RAG pipeline of Mondlock in view of Engi wherein determining the amount of non-overlapping content of each candidate chunk includes determining a cosine-similarity between each candidate chunk relative to each other candidate chunk, as taught by Sivulka, because cosine similarity is one of a finite number of identified, predictable similarity metrics with a reasonable expectation of success, and thus would have been obvious to try.
Claims 4, 5, 11, 12, 18, and 19 are rejected under 35 U.S.C. 103 as being unpatentable over Mondlock et al. (US Patent No. 12,079,570) in view of Engi et al. (US Publication No. 2025/0371066) as applied to claims 1, 2, 6-9, 13-16, and 20 above, and further in view of Wong (NPL entitled “String Matching With FuzzyWuzzy,” dated October 27, 2020).
As to claim 4, the limitations of parent claim 1 have been discussed above. Mondlock in view of Engi does not specifically disclose wherein determining the amount of non-overlapping content of each candidate chunk includes determining a token sort ratio between tokens of each candidate chunk relative to each other candidate chunk. However, Wong teaches
wherein determining the amount of non-overlapping content of each candidate chunk [string] includes determining a token sort ratio between tokens of each candidate chunk relative to each other candidate chunk (see e.g., p. 1 for Fuzzywuzzy using a similarity ratio between two sequences and returning the similarity percentage and p. 3 for FuzzyWuzzy having token functions that tokenize the strings, change capitals to lowercase, and remove punctuation, the token_sort_ratio() function sorting the strings alphabetically and then joining them together, then, the fuzz.ratio() being calculated, and this coming in handy when the strings you are comparing are the same in spelling but are not in the same order).
It would have been obvious to one of ordinary skill in the art prior to the effective filing date of the claimed invention to modify the RAG pipeline of Mondlock in view of Engi wherein determining the amount of non-overlapping content of each candidate chunk includes determining a token sort ratio between tokens of each candidate chunk relative to each other candidate chunk, as taught by Wong, for the benefit of comparing strings with the same spelling but not the same order (see e.g., Wong, p. 3).
As to claim 5, the limitations of parent claim 1 have been discussed above. Mondlock in view of Engi does not specifically disclose wherein determining the amount of non-overlapping content of each candidate chunk includes determining a token set ratio between tokens of each candidate chunk relative to each other candidate chunk. However, Wong teaches
wherein determining the amount of non-overlapping content of each candidate chunk [string] includes determining a token set ratio between tokens of each candidate chunk relative to each other candidate chunk (see e.g., p. 1 for Fuzzywuzzy using a similarity ratio between two sequences and returning the similarity percentage and p. 3 for the token_set_ratio() function being similar to the token_sort_ratio() function above, except it taking out the common tokens before calculating the fuzz.ratio() between the new strings and this function being the most helpful when applied to a set of strings with a significant difference in lengths).
It would have been obvious to one of ordinary skill in the art prior to the effective filing date of the claimed invention to modify the RAG pipeline of Mondlock in view of Engi wherein determining the amount of non-overlapping content of each candidate chunk includes determining a token set ratio between tokens of each candidate chunk relative to each other candidate chunk, as taught by Wong, for the benefit of comparing strings with a significant difference in lengths (see e.g., Wong, p. 3).
As to claim 11, the limitations of parent claim 8 have been discussed above. Mondlock in view of Engi does not specifically disclose wherein determining the amount of non-overlapping content of each candidate chunk includes determining a token sort ratio between tokens of each candidate chunk relative to each other candidate chunk. However, Wong teaches
wherein determining the amount of non-overlapping content of each candidate chunk [string] includes determining a token sort ratio between tokens of each candidate chunk relative to each other candidate chunk (see e.g., p. 1 for Fuzzywuzzy using a similarity ratio between two sequences and returning the similarity percentage and p. 3 for FuzzyWuzzy having token functions that tokenize the strings, change capitals to lowercase, and remove punctuation, the token_sort_ratio() function sorting the strings alphabetically and then joining them together, then, the fuzz.ratio() being calculated, and this coming in handy when the strings you are comparing are the same in spelling but are not in the same order).
It would have been obvious to one of ordinary skill in the art prior to the effective filing date of the claimed invention to modify the RAG pipeline of Mondlock in view of Engi wherein determining the amount of non-overlapping content of each candidate chunk includes determining a token sort ratio between tokens of each candidate chunk relative to each other candidate chunk, as taught by Wong, for the benefit of comparing strings with the same spelling but not the same order (see e.g., Wong, p. 3).
As to claim 12, the limitations of parent claim 8 have been discussed above. Mondlock in view of Engi does not specifically disclose wherein determining the amount of non-overlapping content of each candidate chunk includes determining a token set ratio between tokens of each candidate chunk relative to each other candidate chunk. However, Wong teaches
wherein determining the amount of non-overlapping content of each candidate chunk [string] includes determining a token set ratio between tokens of each candidate chunk relative to each other candidate chunk (see e.g., p. 1 for Fuzzywuzzy using a similarity ratio between two sequences and returning the similarity percentage and p. 3 for the token_set_ratio() function being similar to the token_sort_ratio() function above, except it taking out the common tokens before calculating the fuzz.ratio() between the new strings and this function being the most helpful when applied to a set of strings with a significant difference in lengths).
It would have been obvious to one of ordinary skill in the art prior to the effective filing date of the claimed invention to modify the RAG pipeline of Mondlock in view of Engi wherein determining the amount of non-overlapping content of each candidate chunk includes determining a token set ratio between tokens of each candidate chunk relative to each other candidate chunk, as taught by Wong, for the benefit of comparing strings with a significant difference in lengths (see e.g., Wong, p. 3).
As to claim 18, the limitations of parent claim 15 have been discussed above. Mondlock in view of Engi does not specifically disclose wherein determining the amount of non-overlapping content of each candidate chunk includes determining a token sort ratio between tokens of each candidate chunk relative to each other candidate chunk. However, Wong teaches
wherein determining the amount of non-overlapping content of each candidate chunk [string] includes determining a token sort ratio between tokens of each candidate chunk relative to each other candidate chunk (see e.g., p. 1 for Fuzzywuzzy using a similarity ratio between two sequences and returning the similarity percentage and p. 3 for FuzzyWuzzy having token functions that tokenize the strings, change capitals to lowercase, and remove punctuation, the token_sort_ratio() function sorting the strings alphabetically and then joining them together, then, the fuzz.ratio() being calculated, and this coming in handy when the strings you are comparing are the same in spelling but are not in the same order).
It would have been obvious to one of ordinary skill in the art prior to the effective filing date of the claimed invention to modify the RAG pipeline of Mondlock in view of Engi wherein determining the amount of non-overlapping content of each candidate chunk includes determining a token sort ratio between tokens of each candidate chunk relative to each other candidate chunk, as taught by Wong, for the benefit of comparing strings with the same spelling but not the same order (see e.g., Wong, p. 3).
As to claim 19, the limitations of parent claim 15 have been discussed above. Mondlock in view of Engi does not specifically disclose wherein determining the amount of non-overlapping content of each candidate chunk includes determining a token set ratio between tokens of each candidate chunk relative to each other candidate chunk. However, Wong teaches
wherein determining the amount of non-overlapping content of each candidate chunk [string] includes determining a token set ratio between tokens of each candidate chunk relative to each other candidate chunk (see e.g., p. 1 for Fuzzywuzzy using a similarity ratio between two sequences and returning the similarity percentage and p. 3 for the token_set_ratio() function being similar to the token_sort_ratio() function above, except it taking out the common tokens before calculating the fuzz.ratio() between the new strings and this function being the most helpful when applied to a set of strings with a significant difference in lengths).
It would have been obvious to one of ordinary skill in the art prior to the effective filing date of the claimed invention to modify the RAG pipeline of Mondlock in view of Engi wherein determining the amount of non-overlapping content of each candidate chunk includes determining a token set ratio between tokens of each candidate chunk relative to each other candidate chunk, as taught by Wong, for the benefit of comparing strings with a significant difference in lengths (see e.g., Wong, p. 3).
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Parab et al. (US Publication No. 2017/0090786) for “The system 125, in effect, narrows the comparison down to the most likely candidate chunks or extents stored in the data store 105. In some instances, the system 105 may select extents by comparing root (or head) signatures for a chunk of an input data stream to root (or head) signatures of extents stored in the data store 105. Extents that have matching signatures may be ignored as the blocks corresponding thereto are already present. This process is known as deduplication. That is, only unique data need be transmitted and stored after its identification” (see [0051]).
Any inquiry concerning this communication or earlier communications from the examiner should be directed to DARA J GLASSER whose telephone number is (571)270-3666. The examiner can normally be reached Monday-Thursday, 10:00am-2:00pm.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Apu Mofiz can be reached at (571)272-4080. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
03-08-2026
/DARA J GLASSER/Examiner, Art Unit 2161
/APU M MOFIZ/Supervisory Patent Examiner, Art Unit 2161