DETAILED ACTION
Claims 1 – 20 are pending.
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 .
Information Disclosure Statement
The information disclosure statement (IDS) submitted on 28 June 2024 is in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statement is being considered by the Examiner.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 1 – 20 are rejected under 35 U.S.C. 101 because the claimed invention is directed to a judicial exception without significantly more.
Claims 1 and 11 provide teaching for inputting a first prompt containing a content portion and a plurality of user personas into a first machine learning model, applying the machine learning model and a first prompt to generate a plurality of points of interests associated with the user personas and the content portion, getting a second prompt that includes mappings of the plurality of points of interests to question types and inputting the second prompt into a second machine learning model, the second machine learning model then generating a plurality of questions associated with the user personas and the content portion, and then retrieving a second content portion based on the plurality of questions and a third prompt.
Nothing in the claims precludes them from being performed in the human mind. The entire process involves data gathering involving the inputting of the first prompt, the second prompt, and data generation involving the generation of points of interests, the plurality of questions, the second content portion. The entire process can be performed by a human who receives a first document portion that contains a first portion of a document and different tones or emotions (as personalities to apply), applying a thought process to the first prompt to generate different topics associated with the different personalities and the first document portion, getting a second prompt that now has an association of the different topics and certain question types, applying a thought process to this second prompt to then come up with a plurality of questions that are associated with the personalities and the first document portion, and finally applying or answering the plurality of questions as well as the use of a third prompt to obtain a second document portion. The mentioning of the at least one processor simply serves as available hardware to be able to perform the claimed invention. The claims here recite a mental process.
Claim 19 provides teaching for a processor being used to evaluate a RAG pipeline by using source data and synthetically generated question variants, using machine learning algorithms to generate a plurality of initial questions by applying the source data and the persona, and then making use of one or more language models to process the initial questions and the persona to generate the synthetically generated questions.
Nothing in the claim precludes it from being performed in the human mind. The entire process involves data generation by producing the initial questions and the synthetically generated question, as well as data analysis based on the use of the source data and the persona information. The entire process can be performed by a human who receives a persona information and a source document, applies reasoning on the source document and the persona to generate questions, as well as generating further variations of the previously generated questions. The mentioning of the at least one processor simply serves as available hardware to be able to perform the claimed invention. The machine learning models here appears as a pre-solution activity, which can be performed as reasoning, by a human. The claim briefly mentions evaluating a RAG without particular details about it, and a human may mentally perform this by searching through external data sources to address an input query and obtaining relevant information, augmenting the original query with query variations based on the obtained relevant information, and generating suitable results, which then get evaluated to understand he relevancy of the results to the initial query. The claim here recites a mental process.
This judicial exception is not integrated into a practical application as the claims simply teach of gathering data, analysing data, and generating data.
The invention is not tied to any particular defining structure and simply process instructions to apply the judicial exception. The techniques can be performed by a generic computer which would be presented as a tool to implement the abstract idea (classifiable as automation of a mental concept). The Specification in [0180] provides the use of a generic computer, suitable to execute the claimed technique. The machine learning models presented here can be represented as large language models or any other machine learning model presented in [0054] and [0058] of the Specification. The claims do not present any information regarding the training of the machine learning models, or the particular techniques in which the machine learning models are being applied to the generation of the points of interests or the plurality of questions. By this presentation, mentioning the machine learning models in the claims simply refer to a random or general-purpose machine learning model being applied to perform a mental process. The machine learning model are presented in a way that depict a human performing a reasoning task to be able to generate points of interest from a document while placing one’s self in a particular personality, and for generating questions applicable to that personality. These as can be seen, can be performed mentally. The claims do not include additional elements that are sufficient to amount to significantly more than the judicial exception because the invention is not tied to a practical application.
The claims provide techniques that amount to no more than mere instructions that apply the judicial exception which can be performed by a generic device. Merely mentioning the processor amounts to no more than a general-purpose hardware used as a tool to implement the abstract idea and does not provide any particular application other than applying it for the purpose of implementing a judicial exception. While the claims mention machine learning models, the machine learning models do not recite specifics on how they are being trained nor how they are being applied to the tasks they are presented to perform, and therefore, the claims still do not amount to significantly more than the mentioned judicial exception. Mere instructions to apply an exception using a generic device cannot provide an inventive concept. Claims 1, 11 and 19 are not eligible.
For claims 2 and 12, a human may perform this by collecting the points of interest and the question types to generate the plurality of mappings between the points of interest and the question types. The presence of the third machine learning model simply is provided as a tool through which a human may perform by reasoning and generating a map between datapoints. This does not integrate any practical application nor does it provide any additional element sufficient to amount to more than the mentioned judicial exception.
For claim 3, a human may generate a set of clusters for the points of interest and sort out duplicates of the points of interest based on the set of clusters, before performing the task of mapping. This does not integrate any practical application nor does it provide any additional element sufficient to amount to more than the mentioned judicial exception.
For claim 4, it is provided that the set of clusters is generated based on a plurality of embeddings of the points of interest, which is a mental task that a human may perform by observing the different points of interest. This does not integrate any practical application nor does it provide any additional element sufficient to amount to more than the mentioned judicial exception.
For claim 5, a human may filter the questions based on either of semantic representations, relevance of the questions to the portion of content, tones of the questions, or levels of nuance of the questions. This presents a mental task that can be performed by a human. This does not integrate any practical application nor does it provide any additional element sufficient to amount to more than the mentioned judicial exception.
For claims 6 and 13, a human may convert the questions into variants of the questions that can be associated with different personas. The presence of the third machine learning model simply is provided as a tool through which a human may perform by reasoning and generating a different way to ask a question. This does not integrate any practical application nor does it provide any additional element sufficient to amount to more than the mentioned judicial exception.
For claim 7, a human may first prompt the system to generate the different points of interest by reasoning through a certain way, and have the second prompt to be to generate questions by reasoning through a different other way. The presence of the reasoning structures simply serves the task of reasoning or thinking through a certain way. This does not integrate any practical application nor does it provide any additional element sufficient to amount to more than the mentioned judicial exception.
For claim 8, a human may learn parameters of a model based on data that includes the questions paired with the first document portion, generate a representation of the third prompt and another representation of the second document portion, and retrieve the second document portion based on the two obtained representations. This does not integrate any practical application nor does it provide any additional element sufficient to amount to more than the mentioned judicial exception.
Claims 9 and 17 provide that the first machine learning model can be one of an LLM, VLM or a multi-modal language model, providing simply, the type of machine learning model being used. This does not integrate any practical application nor does it provide any additional element sufficient to amount to more than the mentioned judicial exception.
Claim 10 provides that the persona can comprise at least one of a name, role, etc., providing simply, the possible personas a human can account for. This does not integrate any practical application nor does it provide any additional element sufficient to amount to more than the mentioned judicial exception.
For claim 14, a human may generate a set of clusters for the different questions and sort out duplicates of the questions based on the set of clusters, before performing the task of mapping. This does not integrate any practical application nor does it provide any additional element sufficient to amount to more than the mentioned judicial exception.
For claim 15, a human may generate a representation of the third prompt and another representation of the second document portion, and retrieve the second document portion based on the two obtained representations. This does not integrate any practical application nor does it provide any additional element sufficient to amount to more than the mentioned judicial exception.
For claim 16, a human may evaluate the performance of an embedding model based on two other embedding representations, and this can be performed mentally as a mathematical process. This does not integrate any practical application nor does it provide any additional element sufficient to amount to more than the mentioned judicial exception.
Claim 18 simply lists various systems that the claimed invention could be applied to. This does not integrate any practical application nor does it provide any additional element sufficient to amount to more than the mentioned judicial exception.
Claim 20 simply lists various systems that the claimed invention could be applied to. This does not integrate any practical application nor does it provide any additional element sufficient to amount to more than the mentioned judicial exception.
Claim Rejections - 35 USC § 102
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (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 the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(1) the claimed invention was patented, described in a printed publication, or in public use, on sale, or otherwise available to the public before the effective filing date of the claimed invention.
Claims 1, 10 and 11 are rejected under 35 U.S.C. 102(a)(1) as being anticipated by Shekhar et al. (US 2023/0351096 A1: hereafter — Shekhar).
For claim 1, Shekhar discloses a method comprising:
inputting a first prompt that includes (i) a first portion of content and (ii) a plurality of user personas into a first machine learning model (Shekhar: [0017] — a persona-specific navigation interface that allows a user specify personas (e.g. CEO, CFO, investor, employee, customers); [0023] — receiving a document as input (the persona and the document corresponding to the claimed first prompt));
generating, via execution of the first machine learning model and based at least on the first prompt, a plurality of points of interest associated with the plurality of user personas and the first portion of content (Shekhar: [0071] — applying a machine learning model to identify an interest within a portion of a document (for the specified persona); [0016] — on a single document, the CEO [persona] may be interested in a portion different from the CFO [persona]; [0017] — ‘[t]he portions of content described in the persona-specific navigation interface have identified interests that map to the one or more interests associated with the selected persona’);
inputting a second prompt that includes a plurality of mappings between the plurality of points of interest and a plurality of question types into a second machine learning model (Shekhar: [0040] — a question generator which generates questions based on paragraphs of the documents (based on the points of interest); [0043]–[0044] — questions also being generated based on specific domains and by target qualities that are sought in the final generated domain-specific questions (these domains and target qualities are akin to the question-type being applied to generate questions));
generating, via execution of the second machine learning model and based at least on the second prompt, a plurality of questions associated with the plurality of user personas and the first portion of content (Shekhar: [0005] — automatically generated questions that are linked to document portions; [0019] — questions which the user may have (being linked to the persona) which are generated and are to be answered by the portions of the document (as a plurality of questions associated with the persona); [0041] — the question generation employs an encoder-decoder architecture (this being a machine learning model)); and
retrieving a second portion of content based at least on the plurality of questions and a third prompt (Shekhar: [0019] — applying the questions and the document to link to a portion of the document, a possible third prompt being such as ‘How much did the CEO earn in 2022?’; [0016] — the navigation interface then presenting the portions of document that are of highest interest (as the retrieval and display of the second portion of content)).
For claim 10, claim 1 is incorporated and the reference of Shekhar discloses the method, wherein the plurality of user personas comprises at least one of a name, a role, a behavioral trait, an emotion, a demographic attribute, a communication style, a level of knowledge, a level of education, an attitude, a motivation, an interest, or a goal (Shekhar: [0016], [0017] — personas as CEO, CFO, Investor, employee, customers (as roles)).
As for claim 11, processor claim 11 and method claim 1 are related as processor and the method of using same, with each claimed element’s function corresponding to the claimed method step. Shekhar in FIG. 10 and in [0098] provide a processor necessary to read upon the limitations of this claim. Accordingly, claim 11 is similarly rejected under the same rationale as applied above with respect to method claim 1.
The following is a quotation of the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(2) the claimed invention was described in a patent issued under section 151, or in an application for patent published or deemed published under section 122(b), in which the patent or application, as the case may be, names another inventor and was effectively filed before the effective filing date of the claimed invention.
Claims 19 and 20 are rejected under 35 U.S.C. 102(a)(2) as being anticipated by Chakraborty et al. (US 12,306,828 B1: hereafter — Chakraborty).
For claim 19, the reference of Chakraborty discloses a system comprising:
one or more processors to evaluate a retrieval augmented generation (RAG) pipeline using source data and a plurality of synthetically generated question variants, wherein a plurality of initial questions are generated based at least on processing the source data and persona data using one or more machine learning models, and the plurality of synthetically generated question variants are generated based at least on one or more language models processing the plurality of initial questions and the persona data (Chakraborty: Col 3 lines 33–59 — a retrieval augmented generation based prompting strategies which uses searchable sources (source data); Col 13 lines 56 — generating variations of a user query (synthetically generated question variants)in order to prompt an LLM (the language model, which is a machine learning model) wherein the prompt can vary based on a persona).
For claim 20, claim 19 is incorporated and the reference of Chakraborty discloses the system, wherein the system is comprised in at least one of:
a system for performing simulation operations;
a system for performing digital twin operations;
a system for performing collaborative content creation for 3D assets;
a system for performing one or more deep learning operations;
a system implemented using an edge device;
a system for generating or presenting at least one or virtual reality content, augmented reality content, or mixed reality content;
a system implemented using a robot;
a system for performing one or more conversational AI operations (Chakraborty: Col 4 lines 63–64 — providing conversational AI);
a system implemented using one or more large language models (LLMs) (Chakraborty: Col 4 lines 40 – 41 — generative AI services including LLM);
a system implementing one or more vision language models (VLMs);
a system implementing one or more multi-modal language models;
a system for generating synthetic data;
a system for performing one or more generative AI operations (Chakraborty: Col 4 lines 40 – 41 — generative AI services);
a system incorporating one or more virtual machines (VMs) (Chakraborty: Col 10 lines 54 – 7 — implementation of the server over virtual systems);
a system implemented at least partially in a data center; or
a system implemented at least partially using cloud computing resources (Chakraborty: Col 9 lines 41–43 — implementation in a cloud environment).
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 2 and 12 are rejected under 35 U.S.C. 103 as being unpatentable over Shekhar (US 2023/0351096 A1) as applied to claims 1 and 11, in view of Saha et al. (US 2024/0371380: hereafter — Saha).
For claim 2, claim 1 is incorporated but the reference Shekhar fails to disclose the limitations of this claim, for which the reference of Saha is now introduced to teach as the method, further comprising:
inputting a fourth prompt that includes (i) the plurality of points of interest and (ii) the plurality of question types into a third machine learning model (Saha: [0046] — a machine learning such as an NLP which is able to classify conversational segments (points of interest of documents) into different categories (taking the categories as domains applied here as question types)); and
generating, via execution of the third machine learning model and based at least on the fourth prompt, the plurality of mappings between the plurality of points of interest and the plurality of question types (Saha: [0046] — a machine learning such as an NLP which is able to classify conversational segments (points of interest of documents) into different categories (taking the categories as domains applied here as question types) (the classifying of the segments into different categories is taken as the mapping of the points of interest with question types)).
The reference of Shekhar provides teaching for obtaining a plurality of points of interest as attributable to a persona and a document, and then applying these to generate questions. This however differs from the claimed invention in that the claimed invention further provides teaching for a machine learning model being applied to receive as a fourth prompt, the points of interest question types to be able to determine mappings between them. This is however not new to the art as the reference of Saha is introduced to teach above.
Hence, before the effective filing date of the claimed invention, one of ordinary skill in the art would have found it obvious to combine the known teaching of Saha which classifies conversational segments as points of interest in a document into particular categories, with the teaching of the obtaining a plurality of points of interest as attributable to a persona and a document, and then applying these to generate questions as taught by Shekhar, to thereby come up with the claimed invention. The combination of both prior art elements would have provided the predictable result of ensuring that the points of interest are properly mapped to the categories they best fit into, so that the right questions attributable to such points of interest can be generated, leading to knowing categorical questions to be generated. See KSR Int’l Co. v. Teleflex Inc., 550 U.S. 398, 415-421, 82 USPQ2d 1385,1395-97 (2007).
As for claim 12, processor claim 12 and method claim 2 are related as processor and the method of using same, with each claimed element’s function corresponding to the claimed method step. Accordingly, claim 12 is similarly rejected under the same rationale as applied above with respect to method claim 2.
Claim 3 is rejected under 35 U.S.C. 103 as being unpatentable over Shekhar (US 2023/0351096 A1) in view of Saha (US 2024/0371380) as applied to claim 2, further in view of Wang et al. (US 2020/0137512 A1: hereafter — Wang).
For claim 3, claim 2 is incorporated and the combination of Shekhar in view of Saha discloses the method, further comprising:
generating a set of clusters associated with the plurality of points of interest (Shekhar [0053] — grouping of interests within a document); and
[[deduplicating the plurality of points of interests based at least on the set of clusters]] prior to inputting the plurality of points of interest into the third machine learning model (Saha: [0046] — a machine learning such as an NLP which is able to classify conversational segments (points of interest of documents) into different categories (taking the categories as domains applied here as question types, thereby teaching of the individual inputting of the points of interest into the [third] machine learning model)).
The combination of Shekhar in view of Saha however fails to disclose the further limitation of this claim, for which the reference of Wang is now introduced to teach as:
deduplicating the plurality of points of interests based at least on the set of clusters [[prior to inputting the plurality of points of interest into the third machine learning model]] (Wang: [0070] — identifying duplicate POIs from a first set (cluster) of POIS and a second set (cluster) of POIs, and then deduplicating the POIs that are found to have been duplicated).
The combination of Shekhar in view of Saha provides teaching for generating clusters of points of interest within a document as well as the inputting the points of interest into a machine learning model, but differs from the claimed invention in that the claimed invention further provides teaching for deduplicating the plurality of the points of interest based on sets of clusters. This isn’t new to the art as the reference of Wang goes to teach above.
Hence, before the effective filing date of the claimed invention, one of ordinary skill in the art would have found it obvious to improve upon the teaching of the combination of Shekhar in view of Saha which teaches of generating clusters of points of interest within a document as well as the inputting the points of interest into a machine learning model, by incorporating the known teaching of Wang which deduplicates the plurality of the points of interest based on sets of cluster, to thereby come up with the claimed invention. While the reference of Wang has its points of interest being directed to geographic locations, the points of interests are similar in the sense that they are insights which are considered to be of interest to particular users or personas. The combination of both prior art elements would have provided the predictable result of reducing redundancy by not having to process the same information multiple times. See KSR Int’l Co. v. Teleflex Inc., 550 U.S. 398, 415-421, 82 USPQ2d 1385,1395-97 (2007).
Claim 4 are rejected under 35 U.S.C. 103 as being unpatentable over Shekhar (US 2023/0351096 A1) in view of Saha (US 2024/0371380), further in view of Wang (US 2020/0137512 A1) as applied to claim 3, and further in view of Rollings et al. (US 2021/0382927 A1: hereafter — Rollings).
For claim 4, claim 3 is incorporated but the combination of Shekhar in view of Saha further in view of Wang fails to disclose the limitation of this claim, for which the reference of Rollings is now introduced to teach as the method, wherein the set of clusters is generated based at least on a plurality of embeddings of the plurality of points of interest (Rollings: [0017] — extracting pertinent portions (portions of interest) from a document; [0021] — generating embeddings of the portions to generate portion embedding vectors and then clustering the portions based on the portion embedding vectors).
The combination of Shekhar in view of Saha further in view of Wang provides teaching for obtaining a set of clusters of the points of interest, but differs from the claimed invention in that the claimed invention further provides teaching for obtaining the clusters based on a plurality of embeddings of the points of interest. This isn’t new to the art as the reference of Rollings is seen to teach this above.
Hence, before the effective filing date of the claimed invention, one of ordinary skill in the art would have found it obvious to improve upon the teaching of the combination of Shekhar in view of Saha further in view of Wang which obtains a set of clusters of the points of interest, by incorporating the known teaching of Rollings which teaches of obtaining the clusters based on a plurality of embeddings of the points of interest, to thereby come up with the claimed invention. The combination of both prior art elements would have provided the predictable result of easily making comparisons for groupings, based vector similarities or distance between vectors. See KSR Int’l Co. v. Teleflex Inc., 550 U.S. 398, 415-421, 82 USPQ2d 1385,1395-97 (2007).
Claims 5 are rejected under 35 U.S.C. 103 as being unpatentable over Shekhar (US 2023/0351096 A1) as applied to claim 1 in view of AGRAWAL et al. (US 20220327287 A1: hereafter — Agrawal).
For claim 5, claim 1 is incorporated but the reference of Shekhar fails to disclose the limitations of this claim, for which the reference of Agrawal is now introduced to teach as the method, further comprising filtering the plurality of questions based at least on at least one of semantic representations of the plurality of questions, relevances of the plurality of questions to the first portion of content, tones associated with the plurality of questions, or levels of nuance associated with the plurality of questions (Agrawal: [0046] — a filtering mechanism that can re-rank candidate questions based on semantic similarities).
The reference of Shekhar provides teaching for obtaining questions, but differs from the claimed invention in that the claimed invention further provides the filtering of the questions according semantic representations. This is however not new to the art as the reference of Agrawal is seen to teach this above.
Hence, before the effective filing date of the claimed invention, one of ordinary skill in the art would have found it obvious to combine the known teaching of Agrawal which teaches of filtering the plurality of questions according to semantic similarities, with the teaching of Shekhar which provides teaching for generating the plurality of questions, to thereby come up with the claimed invention. The combination of both prior art elements would have provided the predictable result of making use of only the questions that easily making comparisons for groupings, based vector similarities or distance between vectors. See KSR Int’l Co. v. Teleflex Inc., 550 U.S. 398, 415-421, 82 USPQ2d 1385,1395-97 (2007).
Claims 6 and 13 are rejected under 35 U.S.C. 103 as being unpatentable over Shekhar (US 2023/0351096 A1) as applied to claims 1 and 11 in view of Fejes et al. (US 2020/0228476 A1: hereafter — Fejes).
For claim 6, claim 1 is incorporated but the reference of Shekhar fails to disclose the limitation of this claim, for which the reference of Fejes is now introduced to teach as the method, wherein the generating the plurality of questions comprises converting, via execution of a third machine learning model, each question included in the plurality of questions into a plurality of question variants associated with the plurality of user personas (Fejes: [0008] — retrieving a persona, generating a query based on the persona, and generating variations of the query-based on the persona; [0006] — query generation using conversation multiplier (which is a machine learning technique)).
The reference of Shekhar provides teaching for generating a plurality of questions, but differs from the claimed invention further provides teaching for the generation of the plurality of questions being the machine learning generation of question variants based on the different personas. This is however not new to the art as the reference of Fejes is seen to teach above.
Hence, before the effective filing date of the claimed invention, one of ordinary skill in the art would have found it obvious to modify the teaching of Shekhar which is seen to generate a plurality of questions, through the incorporation of the reference of Fejes which applies a machine learning technique to generate variations of a query based on persona, to thereby come up with the claimed invention. The combination of both prior art elements would have provided the predictable result of generating query variations results in providing improvement to the quality of overall query responses, further providing training data to handle possible future queries. See KSR Int’l Co. v. Teleflex Inc., 550 U.S. 398, 415-421, 82 USPQ2d 1385,1395-97 (2007).
As for claim 13, processor claim 13 and method claim 6 are related as processor and the method of using same, with each claimed element’s function corresponding to the claimed method step. Accordingly, claim 13 is similarly rejected under the same rationale as applied above with respect to method claim 6.
Claims 7 rejected under 35 U.S.C. 103 as being unpatentable over Shekhar (US 2023/0351096 A1) as applied to claim 1 in view of ISAACS (US 2025/0327683 A1 further in view of Niu et al. (US 2025/0103592 A1: hereafter — Niu).
For claim 7, the reference of Shekhar fails to explicitly disclose the limitations of this claim.
The reference of Isaacs is introduced to teach as the method, wherein:
the first prompt further includes a first instruction to generate the plurality of points of interest based at least on a first reasoning structure (Isaacs: [0026] — receiving a prompt to generate points of interest (based on map content), this being performed by making use of generative language models as the reasoning structure).
The reference of Shekhar provides teaching for receiving a first prompt including a first content portion and user personas, but differs from the claimed invention in that the claimed invention further provides that the first prompt further includes an instruction to generate the points of interest based on a first reasoning structure. This isn’t new to the art as the reference of Isaacs is seen to teach above.
Hence, before the effective filing date of the claimed invention, one of ordinary skill in the art would have found it obvious to improve upon the teaching of Shekhar which receives a first prompt, by incorporating the teaching of Isaacs which has that the first prompt includes a query to generate a plurality of points of interest through the use of generative language models, to thereby come up with the claimed invention. The combination of both prior art elements would have provided the predictable result of receiving an input query that specifies the intent of the request, enabling easy access to information retrieval. See KSR Int’l Co. v. Teleflex Inc., 550 U.S. 398, 415-421, 82 USPQ2d 1385,1395-97 (2007).
The combination of Shekhar in view of Isaacs provides teaching for receiving a first prompt that requests the generation of points of interest. It differs from the claimed invention in that the claimed invention further provides the following limitation, which the reference of Niu is now introduced to teach as:
the second prompt further includes a second instruction to generate the plurality of questions based at least on a second reasoning structure (Niu: [0068] — prompts to generate questions using neural--network-based language models as the reasoning structure).
Hence, before the effective filing date of the claimed invention, one of ordinary skill in the art would have found it obvious to improve upon the teaching of the combination of Shekhar in view of Isaacs which takes a first prompt that requests the generation of points of interest, by incorporating the teaching of Niu which presents a second prompt that requests the generation of the plurality of questions based on a neural-network-based language model, to thereby come up with the claimed invention. The combination of both prior art elements would have provided the predictable result of presenting specific instructions to generate training data as the questions, these being useful for preparing responses to address possible future queries. See KSR Int’l Co. v. Teleflex Inc., 550 U.S. 398, 415-421, 82 USPQ2d 1385,1395-97 (2007).
Claims 8, 15 and 16 are rejected under 35 U.S.C. 103 as being unpatentable over Shekhar (US 2023/0351096 A1) as applied to claims 1 and 11 in view of Kapoor et al. (US 2019/0087426 A1: hereafter Kapoor).
For claim 8, claim 1 is incorporated but the reference of Shekhar fails to teach the limitations of this claim, for which the reference of Kapoor is now introduced to teach as the method, wherein the retrieving the second portion of content comprises:
updating one or more parameters of an embedding model based at least on training data that includes the plurality of questions paired with the first portion of content (Kapoor: [0044] — the presence of query and page pairs that includes a query joint embedding and a page joint embedding (indicating the pairing of query or question embeddings with embeddings of pages which are portions of documents) as well as updating of the joint embeddings through updating parameters that are the queries/questions and the pages/portions of content; [0017] — training of a joint embedding model);
generating, via the embedding model after the updating, (i) a first embedding of the third prompt (Kapoor: [0027] — ‘[t]he query embedding model can be trained to compute query embeddings in a multi-dimensional word space’ (teaching of computing an embedding for a prompt after having trained and updated the embedding model)) and (ii) a second embedding of the second portion of content (Kapoor: [0027] — training to obtain page embeddings, whereby page embeddings and query embeddings may be mapped into a page-query joint space); and
retrieving the second portion of content based at least on the first embedding and the second embedding (Kapoor: [0026] — the vector representations/embeddings of the query (prompt) and the pages may be compared to find the page (second portion of content) that closely matches the prompt; [0050] — identifying page results for a user query based on both embeddings of the query and the portion of content).
The reference of Shekhar provides teaching for receiving prompts and first content portions, and providing content portions in response. This differs from the claimed invention in that the claimed invention further provides teaching for the updating of an embedding model on question embeddings and first content portion embeddings, to then retrieve second content portions. This isn’t new to the art as it is seen to be taught by the reference of Kapoor above.
Hence, before the effective filing date of the claimed invention, one of ordinary skill in the art would have found it obvious to modify the teaching of Shekhar which receiving prompts and first content portions, and providing content portions in response, by introducing the known teaching of Kapoor which applies trained embedding models updated on question and content portion embeddings to be able to retrieve second content portions that match an input query prompt, with the teaching of Shekhar which receiving prompts and first content portions, and providing second content portions in response, to thereby come up with the claimed invention. The combination of both prior art elements would have provided the predictable result of more easily making comparisons for matching queries to responses based on computing vector similarities by the distance between the vectors. See KSR Int’l Co. v. Teleflex Inc., 550 U.S. 398, 415-421, 82 USPQ2d 1385,1395-97 (2007).
For claim 15, claim 11 is incorporated but the reference of Shekhar fails to disclose the limitations of this claim, for which the reference of Kapoor is now introduced to teach as the at least one processor, wherein retrieving the second portion of content comprises:
generating, via an embedding model and based on the third prompt and the plurality of questions, (i) a first embedding of the third prompt (Kapoor: [0027] — ‘[t]he query embedding model can be trained to compute query embeddings in a multi-dimensional word space’ (teaching of computing an embedding for a prompt after having trained and updated the embedding model)) and (ii) a second embedding of the second portion of content (Kapoor: [0027] — training to obtain page embeddings, whereby page embeddings and query embeddings may be mapped into a page-query joint space); and
retrieving the second portion of content based at least on the first embedding and the second embedding (Kapoor: [0026] — the vector representations/embeddings of the query (prompt) and the pages may be compared to find the page (second portion of content) that closely matches the prompt; [0050] — identifying page results for a user query based on both embeddings of the query and the portion of content).
The same motivation for introducing the Kapoor reference as applied to claim 8 above is applicable here still.
For claim 16, claim 15 is incorporated and the combination of Shekar in view of Kapoor discloses the at least one processor, wherein the operations further comprise determining a performance of the embedding model based on the first embedding and the second embedding (Kapoor: [0044] — a ranking loss function being employed to update the joint embeddings through both the query embeddings and the page/content section embedding (the loss function being an operation for evaluating performance)).
Claims 9, 17 and 18 are rejected under 35 U.S.C. 103 as being unpatentable over Shekhar (US 2023/0351096 A1) as applied to claims 1 and 11, in view of Kuan (US 2024/0362409 A1).
For claim 9, claim 1 is incorporated but the reference of Shekhar fails to properly disclose the limitations of this claim, for which the reference of Kuan is now introduced to teach as the method, wherein the first machine learning model includes a large language model (LLM), a vision language model (VLM), or a multi-modal language model (Kuan: [0075] — a prompt generation engine that takes persona input, these being performed using an LLM).
The reference of Shekhar provides teaching for the presence of a machine learning model, but differs from the claimed invention in that the claimed invention now further teaches that the machine learning model includes an LLM. This isn’t new to the art as the reference of Kuan is seen to teach above.
Hence, before the effective filing date of the claimed invention, one of ordinary skill in the art would have found it obvious the use of the LLM of Kuan as an obvious method to try while making use of the machine learning models of Shekhar, to thereby come up with the claimed invention. The combination of both prior art elements would have provided the predictable result of being able to process large and complex language generation tasks. See KSR Int’l Co. v. Teleflex Inc., 550 U.S. 398, 415-421, 82 USPQ2d 1385,1395-97 (2007).
As for claim 17, processor claim 17 and method claim 9 are related as processor and the method of using same, with each claimed element’s function corresponding to the claimed method step. Accordingly, claim 17 is similarly rejected under the same rationale as applied above with respect to method claim 9.
For claim 18, claim 11 is incorporated and as applied to claims 9 and 17 above, the combination of Shekhar in view of Kuan discloses the at least one processor, wherein the processing circuitry is comprised in at least one of:
a system for performing simulation operations;
a system for performing digital twin operations;
a system for performing collaborative content creation for 3D assets;
a system for performing one or more deep learning operations (Kuan: [0045] — deep learning neural network model);
a system implemented using an edge device;
a system for generating or presenting at least one or virtual reality content, augmented reality content, or mixed reality content;
a system implemented using a robot;
a system for performing one or more conversational AI operations (Kuan: [0112] — conversation with an interface persona);
a system implemented using one or more large language models (LLMs) ((Kuan: [0075] — performance using an LLM);
a system implementing one or more vision language models (VLMs);
a system implementing one or more multi-modal language models;
a system for generating synthetic data;
a system for performing one or more generative AI operations;
a system incorporating one or more virtual machines (VMs);
a system implemented at least partially in a data center; or
a system implemented at least partially using cloud computing resources (Kuon: [0134] — cloud computing resources).
Claim 14 is rejected under 35 U.S.C. 103 as being unpatentable over Shekhar (US 2023/0351096 A1) in view of Fejes (US 2020/0228476 A1) as applied to claim 13, further in view of Markson et al. (US 11,301,630 B1: hereafter — Markson).
For claim 14, claim 13 is incorporated and the combination of Shekhar in view of Fejes discloses the at least one processor, wherein the generating the plurality of questions further comprises:
[[deduplicating the plurality of question based at least on the set of clusters prior to]] inputting the plurality of questions into the third machine learning model (Shekhar: [0109] — machine learning algorithms which utilise the training data to find correlation among identified features (indicating the machine learning processing of already available features); [0084] — word vectors are first clustered before a machine learning model is applied to the generation of persona-specific points of interest in [0081]).
The combination of Shekhar in view of Fejes provides teaching for clustering as well as performing machine learning algorithms on already pre-processed data.
This combination however fails to teach the further limitations of this claim, for which the reference of Markson is now introduced to teach as:
generating a set of clusters associated with the plurality of questions (Markson: Col 6 lines 30–51 — clustering questions); and
deduplicating the plurality of question based at least on the set of clusters prior to inputting the plurality of questions into the third machine learning model (Markson: Markson: Col 7 lines 11–20 — when duplicates of questions are encountered in a cluster, the duplicate may be excluded, or only one of the questions may be applied).
The combination of Shekhar in view of Saha provides teaching for generating clusters of points of interest within a document as well as the inputting the points of interest into a machine learning model, but differs from the claimed invention in that the claimed invention further provides teaching for deduplicating the plurality of the points of interest based on sets of clusters. This isn’t new to the art as the reference of Wang goes to teach above.
Hence, before the effective filing date of the claimed invention, one of ordinary skill in the art would have found it obvious to improve upon the teaching of the combination of Shekhar in view of Fejes which teaches of generating word clusters, these being applied to the determining of points of interest within a document base on machine learning algorithms, by incorporating the known teaching of Markson which generates question clusters and deduplicates the question clusters, to thereby come up with the claimed invention. The combination of both prior art elements would have provided the predictable result of reducing redundancy by not having to process the same information multiple times. See KSR Int’l Co. v. Teleflex Inc., 550 U.S. 398, 415-421, 82 USPQ2d 1385,1395-97 (2007).
Conclusion
The prior art made of record and not relied upon is considered pertinent to Applicant’s disclosure.
Zafar et al. (US 12,155,742 B1) provides teaching for a RAG model which generates text based on both input query and retrieved documents or passages (Col 28 lines 10–15).
DOGGETT et al. (US 2022/0070532 A1) provides teaching for an insight engine that generates insight datasets based on clusters, the insight engine selecting each of training states share by a predetermined number of state paths, and then deduplicating the selected training states to generate an insight dataset [0046].
Wang et al. (US 2020/0137512 A1) provides teaching for the deduplication of points of interest from different sources (Abstract).
RING et al. (US 2022/0027977 A1) provides teaching for generating personality-based questions [0080].
Castiglia et al. (US 2024/0242087 A1) provides teaching for receiving a plurality of embedding components for the purpose of solving an optimisation problem that minimises the weights on the layers of a machine learning model through the plurality of embeddings, to be able to maintain model performance [0026].
Ganesh et al. (US 12,282,504 B1) provides teaching for generating query embeddings for user queries using a trained embedding model, accessing stored knowledge graphs that includes a plurality of nodes representing documents, each node being associated with a document embedding from the plurality of document embeddings that characterise content of a document (Col 16 line 59 – Col 17 line 1), and receiving a user query, generating a query embedding for the user query using the trained embedding model, accessing a vector database that stores a plurality of document embeddings, retrieving content from clusters of nodes and formulating a response to the user query based on the retrieved content (Col 18 lines 4–23).
Any inquiry concerning this communication or earlier communications from the Examiner should be directed to OLUWADAMILOLA M. OGUNBIYI whose telephone number is (571)272-4708. The Examiner can normally be reached Monday – Thursday (8:00 AM – 5:30 PM Eastern Standard 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, PARAS D. SHAH can be reached at (571) 270-1650. 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.
/OLUWADAMILOLA M OGUNBIYI/Examiner, Art Unit 2653
/Paras D Shah/Supervisory Patent Examiner, Art Unit 2653
05/16/2026
Prosecution Insights