DETAILED ACTION
Introduction
This office action is in response to Applicant’s Amended submission filed on 2/19/2026. Applicant has amended independent claims 7, 21 and 28, and cancelled dependent claims 9, and 23. Claims 7-8,10-18, 21-22, and 24-28 are pending and have been examined.
Response to Amendment and Arguments
Claim objections
Applicant’s amendment and argument have been fully considered, and is determined to be persuasive, therefore the objections have been withdrawn.
35 U.S.C. 101 Rejections
Applicant’s amendment and argument have been fully considered, and is determined to be persuasive, therefore the rejection have been withdrawn.
35 U.S.C. 103 Rejections
Applicant’s amendments and arguments are considered but are either unpersuasive or moot in view of the new grounds of rejection that, if presented, were necessitated by the amendments to the Claims.
Applicant’s arguments are directed to material that is added by the most recent amendments to the Claims. Response, p.11.
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 7-8, 11-13, 21-22, 25-26, and 28 are rejected under 35 U.S.C. 103 as being unpatentable over Stratton (US 20240370339), in view of Chao (US 20240403373), further in view of Duan (US 20230359441), and furthermore in view of Cadoni (US 20240354130).
Stratton discloses: A method of data processing, comprising: converting, by a prediction and execution system, a natural language input into a vector based at least in part on using a text embedding function to process one or more tokens in the natural language input; ([0007] The input may be tokenized with some keywords extracted that are used to filter the large of amount of data included the backup data to filter down to a smaller subset of data. [0043] Computing embeddings may include data conversion to text to obtain embeddings to that are based on text.)
retrieving a plurality of chunks from a first datastore based at least in part on using one or more search indexes stored in a second datastore ([0113] The lookup by database layer 320 from index of embeddings 164 may return the relevant paragraphs/chunk of data (if stored in index of embeddings 164) or a pointer/offset to the data's location (which data access layer 540 may obtain and append into the prompt). This cooperation between database layer 320 and data access layer 502 in this way provides an on-demand data payload from a backup or file storage system 115 for use by AI-driven applications, without having to create embeddings in advance for all data stored to storage system 115. Query processor 321 applies model 322 to the prompt and returns an evaluated answer (582). API access layer 504 includes the answer in response 332.) [The described system is a standard Retrieval-Augmented Generation (RAG) pipeline, which inherently operates on this principle: the index is built and stored separately from the main data (or as a specific indexed view), and this index is then used to retrieve the actual relevant content chunks from the main data source.]
generating a prompt that includes tokens from the natural language input, tokens from one or more of the plurality of chunks retrieved from the first datastore, and instructions for generating a response to the natural language input; ([0109] An example prompt for the payload sent to query processor 321 is as follows: ([0110] 1) user_msg= “Answer the user-provided question based only on the context provided question: {question} context: {context}” [0111] 2) The inputted {question}. [0112] 3) The {context} that was obtained from the index of embeddings 164 search. This is the top n chunks of data/text that matched the artificial neural network query.)
transmitting the prompt via an application programming interface (API) gateway between the prediction and execution system and a large language model (LLM); ([0115] In some examples, a user may provide the data to be queried in association with a query. Response generation platform 500 may provide this data to the user via API access layer 502.)
and verifying a response received from the LLM before transmitting the response to the natural language input. ([0113] Query processor 321 applies model 322 to the prompt and returns an evaluated answer (582). API access layer 504 includes the answer in response 332.)
Stratton does not explicitly disclose the feature below.
Chao discloses: to compare the vector and the one or more tokens in the natural language input to vectors and tokens associated with the plurality of chunks; ([0162] At block 1630, the method 1600 may include comparing a first embedding generated by a large language model (LLM) from the search query to second embeddings generated by the LLM for each of the plurality of data listings to determine a respective relevance for each of the plurality of data listings to the search query.)[A search query is a natural language input and an embedding is a type of vector, data listings function as chunks of data, and second embeddings are vectors. Identifying which data listings are most relevant to the query is equivalence of identifying which chunks are most relevant. The text describes a specific implementation of a semantic search, which directly maps to the claim's requirement of comparing input tokens/vectors against stored chunk tokens/vectors to determine relevance.]
Chao also disclose generating a prompt that includes tokens from the natural language input, tokens from one or more of the plurality of chunks retrieved from the first datastore, and instructions for generating a response to the natural language input; ([0130] This information may be provided to the generative engine 514 with a prompt to generate a relevance between the provided search query 502, the data listing 423, and/or information related to the data listing 423D. Utilizing the LLM store 518, the generative engine 514 may generate the listing explanation 916 based on perceived syntactical and/or semantic similarities between the search query 502 and the data listing 423D.) Also see figs 10A/B-12A/B.
Stratton and Chao are considered analogous art. Therefore, 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 Stratton to combine the teaching of Chao, because generative engine may be utilized to improve the embeddings performed by the embedding engine and/or provide an explanation for the recommendations provided by the data listing ranking engine (Chao, [0094]).
Stratton and Chao do not disclose the following features.
Duan discloses: combining a first set of passage identifiers retrieved from a first search index of vector-based objects and a second set of passage identifiers from a second search index of token-based objects into a single ranked list of passages to use for prompt generation; ([0005] The semantically-similar source code segment is retrieved from a retrieval source code database using a hybrid retrieval technique. The retrieval source code database is constructed with equally-sized source code segments from various source code files arranged in the database in the same consecutive order as they appear in the original source code file. The database includes an embedding vector index and a sparse vector index for each source code segment. The hybrid retrieval technique uses an embedding or dense vector and a sparse vector to search for source code segments from the database. The hybrid retrieval technique is based on a term-frequency based retrieval method and an embedding-based retrieval method.) [Although an argument can be made that following prompt generation could be made based on the reference, Examiner will instead rely on the Cadoni reference which is cited below to show that the combination of multiple inputs can be aggregated to generate a prompt.]
Stratton/Chao/Duan are considered analogous art. Therefore, 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 Stratton/Chao to combine the teaching of Duan, because the hybrid retrieval method combines the strength of traditional keyword search with semantic or similarity search to improve the relevance and accuracy of the retrieved information or response (Duan, [0005]).
Stratton/Chao/Duan does not disclose the following features.
Cadoni discloses: wherein the prompt is generated based at least in part on combining the first set of passage identifiers and the second set of passage identifiers; ([0019] Once the sequence termination character has been entered, the writing prompt, the content of any referenced files, and the metadata for the reference files is aggregated to generate a request for writing assistance from the LLM.) Also see para 0025, 0038-0039 [The referenced files and metadata are analogous to passage identifiers mention in the claim. The aggregating corresponds to combining information to generate the prompt. The text teaches a system that takes multiple inputs, including prompt, files, metadata, combines them and sends them to a LLM which reads on the concept of creating a prompt by combining sets of passage references.]
Stratton/Chao/Duan/Cadoni are considered analogous art. Therefore, 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 Stratton/Chao/Duan to combine the teaching of Cadoni, because including contextual content and metadata into prompt would improve relevancy of generated response and decrease chance of AI hallucination (Cadoni, [0020]).
Regarding Claim 8, Stratton/Chao/Duan/Cadoni discloses all of claim 7,
Chao further discloses: ranking the plurality of chunks retrieved from the first datastore by computing a set of vector-based relevancy metrics ([0081] Search results are often ranked based on either popularity, the date of the listing's addition (i.e., “most recent”), alphabetically based on the data listing titles, or a weighted version of the term frequency-inverse document frequency (TF-IDF) (each of these being a distinct option). The TF-IDF is a numerical statistic that is intended to reflect how important a word is to a document in a collection or corpus, and a TF-IDF analysis may result in a score for individual words in a data listing based on how important that word is.) and a set of token-based relevancy metrics for the plurality of chunks. ([0107] In some embodiments, a weighting strategy may be utilized in combining the similarity/distance scores of elements of the data listings 423 in order to appropriately rank, for example, a data listing 423 with one very relevant table and another data listing 423 with more tables that are each slightly less relevant.)
Where the rationale for the combination would be similar to the one provided earlier.
Regarding Claim 11, Stratton/Chao/Duan/Cadoni discloses all of claim 7,
Stratton further discloses: performing a user field access check to verify that a user is authorized to view or access the plurality of chunks before generating the prompt. ([0071] Data access: Data platform 150 may provide role-based access controls (RBAC) for backup data and prevents users from accessing data they don't have permissions for, such as sensitive data (patient data/PII, trade secrets, financials, and more). Response generation platform 158 may in some examples incorporate RBAC, where filter generator 160 generates filter 304 to filter out data that does not align with users' permissions in order to provide responses that do align to users' permissions.)
Regarding Claim 12, Stratton/Chao/Duan/Cadoni discloses all of claim 7,
Stratton further discloses: wherein generating the prompt comprises: masking one or more words or tokens in the natural language input that include personally identifying information (PII) or other sensitive data. ([0071] Data access: Data platform 150 may provide role-based access controls (RBAC) for backup data and prevents users from accessing data they don't have permissions for, such as sensitive data (patient data/PII, trade secrets, financials, and more). Response generation platform 158 may in some examples incorporate RBAC, where filter generator 160 generates filter 304 to filter out data that does not align with users' permissions in order to provide responses that do align to users' permissions.)
Regarding Claim 13, Stratton/Chao/Duan/Cadoni discloses all of claim 7,
Stratton further discloses: wherein the response provided by the LLM contains text extracted from one or more passages and links to the one or more passages provided by the prediction and execution system. ([0107] At step 536, in some aspects, database layer 320 may either save the data/text chunks into embeddings 164 or save a reference link to the data for retrieval.) The disclosure describes a process within a system architecture (a retrieval-augmented generation or RAG system) where the system determines the most efficient way to store and retrieve data/text for later use, supporting the final output provided by an LLM]
Regarding Claim 21, Stratton discloses: An apparatus, comprising: at least one memory storing code; ([0025] Each of storage devices 180 may include system memory. [0119] instructions or code)
and one or more processors coupled with the at least one memory and individually or collectively operable to execute the code to cause the apparatus to: ([0050] One or more processors 213 of computing system 202 may implement functionality and/or execute instructions associated with computing system)
As for the rest of the claim, they recite the method of claim 7, therefore the rejection applied to claim 7 would be similarly applicable.
Regarding Claim 28, Stratton discloses: A non-transitory computer-readable medium storing one or more programs, the one or more programs comprising instructions executable by one or more processors to: ([0119] In one or more examples, the functions described may be implemented in hardware, software, firmware, or any combination thereof. If implemented in software, the functions may be stored, as one or more instructions or code, on and/or transmitted over a computer-readable medium and executed by a hardware-based processing unit.)
As for the rest of the claim, they recite the method of claim 7, therefore the rejection applied to claim 7 would be similarly applicable.
Claim 22 recites limitations similar to the limitations of Claim 8 and is rejected under similar rationale.
Claim 25 recites limitations similar to the limitations of Claim 11 and is rejected under similar rationale.
Claim 26 recites limitations similar to the limitations of Claim 12 and is rejected under similar rationale.
Claims 10 and 24 are rejected under 35 U.S.C. 103 as being unpatentable over Stratton/Chao/Duan/Cadoni and furthermore in view of Yin (US 20160335263).
Regarding Claim 10, Stratton/Chao/Duan/Cadoni discloses all of claim 8,
Stratton/Chao/Duan/Cadoni does not explicitly disclose the below feature.
Yin discloses: wherein ranking the plurality of chunks comprises: removing one or more passages with vector-based relevancy metrics or token- based relevancy metrics below a threshold. ([0048] the search result ranking unit 306 may directly filter or remove a bad result after determining that its score is below the threshold, and then rank the left good results based on their scores.)
Stratton/Chao/Duan/Cadoni/Yin are considered analogous art. Therefore, 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 Stratton/Chao/Duan/Cadoni to combine the teaching of Yin, because filtering and removal of irrelevant result will ensure the user will be provided with higher quality and better response (Yin, [0048]).
Claim 24 recites limitations similar to the limitations of Claim 10 and is rejected under similar rationale.
Claims 14 and 27 are rejected under 35 U.S.C. 103 as being unpatentable over Stratton/Chao/Duan/Cadoni, and furthermore in view of Davis (US 10523643).
Regarding Claim 14, Stratton/Chao/Duan/Cadoni discloses all of claim 7,
Stratton/Chao/Duan/Cadoni does not explicitly disclose the below feature.
Davis discloses: wherein verifying the response comprises: performing a comparison between tokens in the response provided by the LLM, tokens in the natural language input, and tokens in the plurality of chunks retrieved from the first datastore. ([col. 10, lines 54-64] comparison between a user response to a prompt against the information contained in the user profile generated by the content gathering circuit 128. For example, if the user inputs a username into a prompt requesting the same, the security parameter circuit 130 may sweep the user's profile for the inclusion of any information that is related to the username. For example, the user may use the same username for an account at another service provider. Alternatively, the username may be a combination of user attributes (e.g., street name and first name) contained in the user's profile.) [The cited passage describes the concept of comparing tokens in a user's input, an LLM's response, and retrieved data from a database to check for consistency or potential security risks, essentially comparing a user's input against their stored profile information for potential inconsistencies or sensitive data exposure.]
Stratton/Chao/Duan/Cadoni/Davis are considered analogous art. Therefore, 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 Stratton/Chao/Duan/Cadoni to combine the teaching of Davis, because comparing a user's response to a prompt against their existing profile information offers significant benefits for enhanced security, fraud prevention, and improved risk-based authentication (Davis, [col. 10, lines 54-64]).
Claim 27 recites limitations similar to the limitations of Claim 14 and is rejected under similar rationale.
Claim 15 is rejected under 35 U.S.C. 103 as being unpatentable over Stratton/Chao/Duan/Cadoni, and furthermore in view of Holmes (US 20220198028), and furthermore in view of Cline (US 10747894).
Regarding Claim 15, Stratton/Chao/Duan/Cadoni discloses all of claim 7,
Stratton/Chao/Duan/Cadoni does not explicitly disclose the below feature.
Holmes discloses: wherein verifying the response comprises: verifying that a format of the response and any citations therein conform to the instructions provided by the prediction and execution system, ([0044] Further, each of the executed programmatic robots may perform any of the exemplary processes described herein to verify that a structure or format of the corresponding query email message conforms to an expected structure or format, and in response to the verified conformity)
Stratton/Chao/Duan/Cadoni/Holmes are considered analogous art. Therefore, 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 Stratton/Chao/Duan/Cadoni to combine the teaching of Holmes, because checking for proper format ensuring data quality, preventing errors and system failures, and improving overall system efficiency and security (Holmes, [0004]).
Stratton/Chao/Duan/Cadoni/Holmes does not explicitly disclose the following.
Cline discloses: wherein the LLM is configured to delete all data provided by the prediction and execution system after returning the response. ([Abstract] In examples, a sensitivity designation may be received from the third-party application, which may cause the remote system to encrypt the responsive text data, redact the text data, and/or remove the text data from the remote system after the response is provided to the voice-enabled device.)
Stratton/Chao/Duan/Cadoni/Holmes/Cline are considered analogous art. Therefore, 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 Stratton/Chao/Duan/Cadoni/Holmes to combine the teaching of Cline, because proper handling of data privacy and regulatory compliance can be met while system clears out memory and save memory/storage consumption (Cline, [Abstract]).
Claim 16 is rejected under 35 U.S.C. 103 as being unpatentable over Stratton/Chao/Duan/Cadoni, and furthermore in view of Wadhawan (US 20250284886), and Dugan (US 20200242161).
Regarding Claim 16, Stratton/Chao/Duan/Cadoni discloses all of claim 7,
Cadoni further discloses: feedback analysis ([0005] receiving the writing feedback from the AI writing engine; and providing the writing feedback to the writing assistance client.)
Where the rationale for the combination would be similar to the one previously provided.
Stratton/Chao/Duan/Cadoni does not explicitly disclose the below feature.
Wadhawan discloses: wherein verifying the response comprises: analyzing data in the response provided by the LLM for toxicity and bias mitigation, ([0126] a finetuning process of LM 160 to LM 170, and suchlike, enabling a reduction in toxicity/bias/offensiveness,)
Stratton/Chao/Duan/Cadoni/Wadhawan are considered analogous art. Therefore, 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 Stratton/Chao/Duan/Cadoni to combine the teaching of Wadhawan, because toxicity or bias mitigation/reduction would improve end user experience (Wadhawan, [0126]).
Stratton/Chao/Duan/Cadoni/Wadhawan does not explicitly disclose the following features.
Dugan discloses: feedback analysis, and content moderation. ([0048] The content moderation system 140 can receive the rating information 154. As is illustrated in FIG. 2, in some embodiments, the content moderation system 140 can include a feedback component 250 that can receive the rating information 154. In response to receiving the rating information 154, the content moderation system 140 can determine, via the feedback component 250, for example, that a portion of the digital content 148 is classified as non-offensive content.)
Stratton/Chao/Duan/Cadoni/Wadhawan/Dugan are considered analogous art. Therefore, 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 Stratton/Chao/Duan/Cadoni/Wadhawan to combine the teaching of Dugan, because feedback analysis and content moderation would improve end user experience (Dugan, [0048]).
Claim 17 is rejected under 35 U.S.C. 103 as being unpatentable over Stratton/Chao/Duan/Cadoni, and furthermore in view of Felsher (US 8316237).
Regarding Claim 17, Stratton/Chao/Duan/Cadoni discloses all of claim 7,
Stratton/Chao/Duan/Cadoni does not explicitly disclose the below feature.
Felsher discloses: wherein transmitting the prompt to the LLM comprises: establishing a secure communication channel between the prediction and execution system and a provider of the LLM, wherein the prompt and the response are communicated via the secure communication channel. ([Abstract] The data transmission may be made secure with respect to the intermediary by providing an asymmetric key or direct key exchange for encryption of the communication between the first and second party. The data transmission may be made secure with respect to the second party by maintaining the information in encrypted format at the second party, with the decryption key held only by the intermediary, and transmitting a secure composite of the decryption key and a new encryption key to the second party for transcoding of the data record, and providing the new decryption key to the first party, so that the information transmitted to the first party can be comprehended by it.)
Stratton/Chao/Duan/Cadoni/Felsher are considered analogous art. Therefore, 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 Stratton/Chao/Duan/Cadoni to combine the teaching of Felsher, because creating secure channel for data transmission would protect and ensure safety of the data (Felsher, [Abstract]).
Claim 18 is rejected under 35 U.S.C. 103 as being unpatentable over Stratton/Chao/Duan/Cadoni, and furthermore in view of Brancato (US 20230300152).
Regarding Claim 18, Stratton/Chao/Duan/Cadoni discloses all of claim 7,
Stratton further discloses: retaining the prompt and the response in a cache for a ([0114] data access layer 504 is configured to cache all or some portion of a dataset retrieved from storage system 115. Cache 561 represents a storage cache within data access layer 504 that stores, for future queries, data from storage system 115. … cache 561 may be included in the index of embeddings 164 database or other data structure. In addition, index of embeddings 164 may be updated to reference cached data rather than (or in addition to) referencing a location in storage system 115.)
and using one or both of the prompt or the response to process subsequent queries from other users. ([0114] data access layer 504 is configured to cache all or some portion of a dataset retrieved from storage system 115. Cache 561 represents a storage cache within data access layer 504 that stores, for future queries,… Also see claim 14, based on the index of embeddings and the cache, a subsequent query to generate a response for the subsequent query.)
Stratton/Chao/Duan/Cadoni does not explicitly disclose tenant configured retention period.
Brancato discloses: tenant configured retention period. ([0133] The retention period is a user configurable time period for which a cache entry is permitted to be in the detection cache.)
Stratton/Chao/Duan/Cadoni/Brancato are considered analogous art. Therefore, 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 Stratton/Chao/Duan/Cadoni to combine the teaching of Brancato, because enable user/ tenant’s ability to configure retention period allows for great flexibility to suit the needs of the tenant/users (Brancato, [0133]).
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure: Sewark (US 20220414137) – discloses hybrid search involving combination of semantic/keyword search/retrieval.
Thomas (US 20240386058) -discloses generating a prompt using contextual information such as filename, information relating to workbook where the chat is located, number of charts in the spreadsheet or other document, column header information, and etc.
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to Philip H Lam whose telephone number is (571)272-1721. The examiner can normally be reached 9 AM-3 PM Pacific time.
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, Bhavesh Mehta can be reached on 571-272-7453. 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.
/PHILIP H LAM/ Examiner, Art Unit 2656
/BHAVESH M MEHTA/Supervisory Patent Examiner, Art Unit 2656