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 .
Continued Examination Under 37 CFR 1.114
A request for continued examination under 37 CFR 1.114, including the fee set forth in 37 CFR 1.17(e), was filed in this application after final rejection. Since this application is eligible for continued examination under 37 CFR 1.114, and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114.
Applicant's submission filed on 07/29/2026 has been entered.
Status of the Claims
Claims 1, 10, 19 have been amended. Claims 1-20, 22 are pending.
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 1-18 is/are rejected under 35 U.S.C. 103 as being unpatentable over John et al. (US 20150039581) in view of He et al. (US 20240362286).
Regarding claim 1, John teaches a computer-implemented method for searching electronic documents, comprising:
defining a search scope for semantic searching ([0001] “perform a semantic search on a limited set of documents from a document space”, [0034] see “a reduced document space on which to perform the semantic searching operation”) by:
receiving a non-semantic search query from a user to search a document corpus, and executing a non-semantic search according to the non-semantic search query ([0032] “a user to perform keyword searches of patent documents”, [0044]) to generate a first search result that identifies a limited set of first documents from the document corpus ([0034] “enter a keyword or search term for generation of the set of search results that provides a reduced document space on which to perform the semantic searching operation, the GUI may also include an option to select a pre-defined set of documents or to upload a document that includes a list of document identifiers”, [0037] “search set was limited based on user inputs to provide a subset of the document space to the semantic engine for sematic searching”);
storing document identifiers from the non-semantic search result as query parameters for a subsequent search ([0035] “may include a list of previously retrieved sets of documents within which the user may wish to perform the semantic search”, [0036] “retrieve the set of documents corresponding to the user input and provide the set of documents together with the semantic input to semantic search engine”);
receiving a natural language query from the user ([0018] “text inputs for receiving user data”, [0038] “user input includes seed data and a semantic input”)(see NOTE);
servicing the natural language query to generate a response to the user, servicing the natural language query comprising:
executing a semantic search scoped to the limited set of first documents ([0035]-[0036], [0046]) using the stored query parameters to generate a semantic search result that identifies semantically relevant content that is semantically relevant to the natural language query ([0015] “semantic searching within the set of documents to identify a semantic search results that are semantically similar to the semantic input”, [0041]),
wherein executing the semantic search comprises filtering entries in
generating an input
receiving generative text
John does not explicitly teach, however He discloses a vector index ([0203], [0266], [0290]) of semantically embedded text chunks ([0150]-[0151], [0168], [0197]);
generating an input to a Large Language Model (LLM), the input comprising the natural language query and the semantically relevant content from the semantic search result ([0122] “prepare a prompt with both the search query and some or all of the search results (e.g., the top k sections) from the electronic document and send it to the generative AI model”, [0147], [0312]); and
receiving generative text generated by the LLM to respond to the natural language query based on the semantically relevant content ([0147]-[0148], [0205], F17:1706-1712).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the teachings of John to include a vector index and LLM as disclosed by He. Doing so allows for efficient search through and retrieve information from a large corpus of text (He [0131]) and assist in summarizing the search results relevant to a given search query (He [0147]).
NOTE John teaches a user can enter a keyword search by selecting a keyword, category or a predefined input, which construed to be analogous to the “receiving a non-semantic search query from a user.” John further teaches receiving a user input by means of a text input and semantic input, which construed to be analogues “receiving a natural language query from the user.” Still, it is not clear if a search is executed in response to each user input or in response to a combined user input. In one of a different embodiment John teaches a user selecting / uploading a documents to search (seed data – aka first input) – “and including text inputs for receiving the seed data and the semantic input” (see claim 2), wherein it would have been obvious receive the semantic input (aka “a natural language query from the user”) after the execution of the first user query (i.e. separately search the first and second user inputs), because the end result stays the same – a search is performed within the limited set of documents.
However, to merely obviate such reasoning, He discloses receiving and searching separately first and second search input, wherein the first input is lexical and second is semantic and the limitations of -
(I) receiving a non-semantic search query from a user to search a document corpus ([0119], [0131], [0133], [0227], F5:508, F7:508) and
executing a non-semantic search according to the non-semantic search query to generate a first search result that identifies a limited set of first documents ([0076], [0119] “use the lexical search generator to generate a first set of lexical search results 146”, wherein lexical search results 146 is a limited set of first documents) from the document corpus ([0131] “user performs a search for "climate change", the search engine can use the inverted index to quickly retrieve a list of all the articles that contain that term”, [0143] “lexical search generator to perform a lexical full-text search to produce and rank a first set of search results”) and
(II) receiving a natural language query from the user ([0172], [0185] “client may … perform subsequent operations, such as requesting more information about the candidate document … in the search results … a subsequent search query”);
servicing the natural language query to generate a response to the user ([0177], [0183]).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the teachings of John to receive and search separate inputs as disclosed by He. Doing so may allow the user to build search queries in an iterative manner, drilling down on more specific search questions in follow-up to reviewing previous search results (He [0135]).
Claim 10 recites substantially the same limitations as claim 1, and is rejected for substantially the same reasons.
Regarding claims 2 and 11, John as modified teaches the method and the medium, wherein the non-semantic search is a lexical search (He [0029], [0070], [0117], [0119]).
Regarding claims 3 and 12, John as modified teaches the method and the medium, further comprising:
providing the first search result to the user in a graphical user interface; and receiving, via user interaction with the graphical user interface, an indication to scope the semantic search to the first documents (John [0039]-[0041], claim 2, He [0146], [0148], [0228], F15-16).
Regarding claims 4 and 13, He as modified teaches the method and the medium, further comprising automatically scoping the semantic search to the first documents (John [0039], He [0094], [0119]).
Regarding claims 5 and 14, He as modified teaches the method and the medium, wherein the natural language query is a query to an artificial intelligence search assistant (He F15: 1510, 1512).
Regarding claims 6 and 15, He as modified teaches the method and the medium, wherein servicing the natural language query comprises: providing the generative text to the user in response to the natural language query (He F17:1706-1712).
Regarding claims 7 and 16, He as modified teaches the method and the medium, wherein the input to the Large Language Model includes the natural language query as a prompt and the semantically relevant content as a context for responding to the prompt (He [0073], [0122], [0147]-[0148], [0205]).
Regarding claims 8 and 17, He as modified teaches the method and the medium, wherein the semantically relevant content comprises semantically relevant text chunks from the first documents (He [0029], [0031], [0033], John [0041]).
Regarding claims 9 and 18, He as modified teaches the method and the medium, wherein the semantically relevant content comprises semantically relevant documents from the first documents (He [0029], [0031], [0033], John [0040], [0046]).
Claims 19-20, 22 and alternatively Claims 1-18 is/are rejected under 35 U.S.C. 103 as being unpatentable over He et al. (US 20240362286) in view of Mukherjee et al. (US 20240354436) based on the provisional application 63/497,932 (dated 04/24/2023, see attached) or Kritt et al. (US 20130311860) and in further view of John et al. (US 20150039581).
Regarding claim 19, He teaches a computer system providing enhanced search, the computer system comprising:
storage storing: a plurality of snippets, each of the plurality of snippets comprising snippet text extracted from a document in a document corpus and a reference to the document from which the snippet text of that snippet was extracted ([0129]);
an embedding store comprising a vector index of the plurality of snippets ([0041] see “searching a document index of contextualized embeddings for the electronic document with the search vector, where each contextualized embedding comprises a vector representation of a sequence of words in the electronic document that includes contextual information”, [0071], [0116] “document vectors may be indexed and stored as a document index to facilitate search and retrieval operations”, [0130]);
a processor (F20); a non-semantic search engine that is executable to search the document corpus ([0131] “user performs a search for "climate change", the search engine can use the inverted index to quickly retrieve a list of all the articles that contain that term”, [0143] “lexical search generator to perform a lexical full-text search to produce and rank a first set of search results”);
a semantic search engine that is executable to perform semantic searching of the document corpus using the vector index ([0136], [0143] “then use the semantic search … to perform a semantic search that does a semantic re-ranking, which uses the context or semantic meaning of a search query … extracts and returns captions and answers in the response”, [0146]); and
memory storing instructions to:
define a search scope for semantic searching by executing the non-semantic search to generate a first search result that identifies a limited set of first documents from the document corpus ([0143] “substructure returned from a semantic query”, [0183] “search results 146 may comprise a subset of candidate document vectors from the set of candidate document vectors 718”, [0215] “generating a semantic similarity score for each candidate document … and selecting the subset of candidate document vectors”, [0221]);
store document vectors index
service a natural language query received from a user to generate a response to the user ([0177], [0183]), the servicing comprising:
executing a semantic search scoped to the limited set of first documents using the stored index size of data sets, number of electronic documents, compute resources, memory resources, network resources, device resource”, [0135], [0212] “search index … will depend on … desired query performance, and the available system resources”),
wherein executing the semantic search comprises searching entries in the vector index of the plurality of snippets ([0203], [0266], [0290],[0150]-[0151], [0168], [0197]) to only include entries corresponding to the limited set of first documents ([0119] “search manager may use the lexical search generator to generate a first set of lexical search results 146, and the semantic search generator to iterate over the first set of lexical search results 146 to generate a second set of semantic search results 146”; [0135] “search query 144 may be modified or expanded … allow the user to build search queries in an iterative manner, drilling down on more specific search questions in follow-up to reviewing previous search results 146”; [0143] “query execution pipeline in two ways. First, it adds secondary ranking over an initial result set … the search manager may use the lexical search generator to perform a lexical full-text search to produce and rank a first set of search results 146”; “then use the semantic search generator to perform a semantic search that does a semantic re-ranking, which uses the context or semantic meaning of a search query 144 to compute a new relevance score over the first set of search results 146”; [0144] “Using those results as the document corpus, semantic ranking re-scores those results based on the semantic strength of the match”), and performing a semantic similarity search using the
generating an input to a Large Language Model (LLM), the input comprising the natural language query and the semantically relevant content from the semantic search result ([0122] “prepare a prompt with both the search query and some or all of the search results (e.g., the top k sections) from the electronic document and send it to the generative AI model”, [0147]); and
receiving generative text generated by the LLM to respond to the natural language query based on the semantically relevant content ([0147]-[0148], [0205], F17:1706-1712).
He does not explicitly teach, however Mukherjee discloses storing document identifiers (p.5 ¶3.a.ii “Returns the n closest doc ids”) from the non-semantic search result as query parameters for a subsequent search and using the stored query parameters to generate a semantic search result (p.5 ¶4.“Query the ontology with the returned doc ids to pull the document text”, ¶5).
NOTE Mukherjee further discloses search only on the limited set of first documents (p.10 C4).
NOTE I He teaches receiving subsequent queries on the received search results. Given that the initial query is a natural language query, it is reasonable to conclude that the subsequent queries are natural language queries as well.
However, to merely obviate such reasoning, Mukherjee discloses receiving a natural language query from the user ([0006] “receiving natural language prompts and providing natural language responses”); and servicing the natural language query to generate a response to the user (p.5 ¶5 “original question are submitted to an LLM”).
Mukherjee further discloses generating an input to a Large Language Model (LLM), the input comprising the natural language query and the semantically relevant content from the semantic search result (p.9 C1-2); and receiving generative text generated by the LLM to respond to the natural language query based on the semantically relevant content (p.9 C1-2).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the teachings of He to include document identifiers as query parameters and receiving a natural language query from the user as disclosed by Mukherjee. Doing so would ensure that one or more LLMs only provides, for example, responses that are based on permitted information sources (Mukherjee [0004]).
◊ Mukherjee further discloses search only on the limited set of first documents (p.10 C4).
Thus, He teaches using document vectors for subsequent searching. Mukherjee further discloses that such vectors can be document identifiers. Both He and Mukherjee obviously teach the subsequent search is permed on the returned set of documents, as argued above.
However, to merely obviate such reasoning, Kritt discloses –
storing document identifiers from the … search result as query parameters for a subsequent search and executing a … search scoped to the limited set of first documents using the stored query parameters to generate a … search result and search only on the limited set of first documents ([0020]-[0022]).
It would have been obvious to one of ordinary skill in the art at the time of invention to modify the teachings of He to include document identifiers from the non-semantic search result as query parameters to search only on the limited set of first documents as disclosed by Kritt. Doing so provides a convenient way for the user to explore documents which relate to some topic found within a document located by a search (Kritt [0033]).
He does not explicitly teach, however John discloses search comprises filtering entries in the vector index of the plurality of snippets to only include entries corresponding to the limited set of first documents ([0039]-[0041] “semantic search engine searches within the filter set using the semantic input to find the set of semantic search results without rebuilding a semantic index … and the semantic search engine searches within the set of products to find the set of semantic search results without rebuilding a semantic index”).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the teachings of He to include filtering entries in the vector index as disclosed by John. Doing so would allow a searcher to identify a set of semantically relevant search results in a short amount of time (John [0023]).
NOTE II He teaches “lexical search results 146” can be returned in response to a lexical search – where “semantic search generator to iterate over the first set of lexical search results 146 to generate a second set of semantic search results 146” in response to a natural language from the user.
See example in [0131] – a user is searching for “climate change.” A lexical, exact search is performed, such as by user of inverted index and Boolean queries. The set of lexical search results is returned – “listing the article or articles where that term appears. The entry might look something like this: "climate change": article!, article2, article5”. Such results, can reasonably and obviously correspond to “a set of electronic documents 706 to create a set of contextualized embeddings … for document content contained within each electronic document 706” [0123], “encode a set of electronic documents 706 to create a set of contextualized embeddings” [0129]; “receive a search query 144, … to retrieve search results 146 with semantically similar document content within an electronic document 706” [0132].
Given that the user can – “perform lexical searching, semantic searching, or a combination of both” [0036], [0070], it is reasonable to conclude that when the user retrieves a first set of documents by lexical search, as shown in [0131] and then search withing such documents by a semantic search and natural query - “receive a search query to search for information within an electronic document … query may comprise any free form text in a natural language” ([0133]), “client may interact with the GUI view to perform subsequent operations, such as requesting more information about the candidate document vectors in the search results 146, presenting portions of the electronic document 706 containing the candidate document vectors, a subsequent search query 144” [0185]; “perform a semantic search for document content contained within au electronic document 706, and generate a set of search results 146 relevant to the search query” [0200].
He teaches performing “lexical searching, semantic searching, or a combination of both” a single set of search results 146, which can be returned by either lexical or semantic searching – “semantic search generator to iterate over the first set of lexical search results 146 to generate a second set of semantic search results 146”; “use the lexical search … to perform lexical searching in response to a search query 144 … may use the semantic search … to perform semantic searching in response to a search query 144.”
Thus, there is only a single set or search results – 146 and a single query 144. However, a query 144, which can include subsequent queries 144 (see [0185] “a subsequent search query 144”), can be either one - semantic or lexical (defined by the same number 144), which produces either semantic or lexical search results (defined by the same number 146). It is reasonable and obvious to conclude that the subsequent second query 144 (semantic or lexical) which “iterate over the first set of lexical search results 146” can be either one – semantic or lexical, given that the searching can be “a combination of both.” I.e. the search query 144 (which can be semantic or lexical) iterating, searching over search results 146 (which can be semantic or lexical) obviously and reasonably produces all possible combinations of searchings’ – lexical search over the semantic results and a semantic search over the lexical results.
Therefore, it is reasonable and obvious to conclude that given that the semantic query and the lexical query are defined by the same number 144 and a subsequent search query is defined by the same number 144 (which means the subsequent search query 144 can also be semantic or lexical) and the semantic results and the lexical results are defined by the same number 146, produces all possible combinations of searching and is limiting such searching to only search result 146 (which can be semantic or lexical) and satisfies the limitation – “lexical search is constrained to operate only within the first documents identified in the first search result.”
◊ Mukherjee and JOHN further discloses search only on the limited set of first documents (Mukherjee p.10 C4, John [0023], [0036], [0040]-[0041]). Thus, He teaches using document vectors for subsequent searching. Mukherjee further discloses that such vectors can be document identifiers. Thus, He, John and Mukherjee obviously teach the subsequent search is permed on the returned set of documents, as argued above.
Claims 1 and 10 recites substantially the same limitations as claim 19, and is rejected for substantially the same reasons.
Regarding claim 20, He as modified teaches the computer system of Claim 19, wherein the non-semantic search engine is a lexical search engine (He [0029], [0070], [0117], [0119).
Regarding claim 22, He as modified teaches the computer system of Claim 19, wherein the memory further stores instructions executable to:
display the generative text to the user (He F17:1706-1712, Mukherjeep.9 C1).
The dependent claims 2-18 are rejected based on the same reasoning as in the first rejection above and are omitted here for the sake of brevity.
◊ Claims 1, 10 and 19 are additionally and/or alternatively rejected under 35 U.S.C. 103 as being unpatentable over He et al. (US 20240362286) in view of John et al. (US 20150039581).
Regarding claims 1, 10 and 19, He teaches a computer-implemented method, non-transitory, computer-readable medium and a computer system as disclosed above with respect to claim 19.
He does not explicitly teach, however John discloses -
storing document identifiers from the non-semantic search result as query parameters for a subsequent search ([0035] “may include a list of previously retrieved sets of documents within which the user may wish to perform the semantic search”, [0036] “retrieve the set of documents corresponding to the user input and provide the set of documents together with the semantic input to semantic search engine”);
executing a semantic search scoped to the limited set of first documents ([0035]-[0036], [0046]) using the stored query parameters to generate a semantic search result that identifies semantically relevant content that is semantically relevant to the natural language query ([0015] “semantic searching within the set of documents to identify a semantic search results that are semantically similar to the semantic input”, [0041]),
semantic search comprises filtering entries in a … index of semantically embedded text chunks to only include entries corresponding to the limited set of first document ([0039]-[0041] “semantic search engine searches within the filter set using the semantic input to find the set of semantic search results without rebuilding a semantic index … and the semantic search engine searches within the set of products to find the set of semantic search results without rebuilding a semantic index”).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the teachings of He to include filtering entries in the vector index as disclosed by John. Doing so would allow a searcher to identify a set of semantically relevant search results in a short amount of time (John [0023]).
Response to Arguments
Applicant's arguments, filed 07/29/2026, in regard to the presently amended claims are addressed in the updated rejections to the claims above.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure is indicated on PTO-892.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to POLINA G PEACH whose telephone number is (571)270-7646. The examiner can normally be reached Monday-Friday, 9:30 - 5:30.
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, Aleksandr Kerzhner can be reached at 571-270-1760. 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.
/POLINA G PEACH/ Primary Examiner, Art Unit 2165 September 22, 2026
Prosecution Insights