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 .
Status of Claims
Claims 1-7, 9-14, and 17-19 are pending of which claims 1, 9 and 16 are in independent form.
Claims 1-7, 9-14, and 17-19 are rejected under 35 U.S.C. 101.
Claims 1-7, 9-14, and 17-19 are rejected under 35 U.S.C. 103.
Response to Arguments
Applicant’s arguments with respect to claim(s) 1-7, 9-14, and 17-19 have been considered but are moot because the new ground of rejection does not rely on any reference applied in the prior rejection of record for any teaching or matter specifically challenged in the argument.
Regarding the 35 USC 101 (Abstract Idea):
Applicant(s) arguments have been fully considered but are not persuasive. Applicant contends that the claims are directed to a specific ML process involving AI agents, query analysis, query reformulation, embedding generation, retrieval of query relevant content, and LLM based response generation. However, the Examiner maintains that the claims recite metal processes. Specifically, determining whether a query is on-topic or off-topic, identifying a key entity from the query, reformulating the query based on the identified entity, and determining a communication type associated with the entity constitute observations, evaluations, classifications, and judgement that can be practically performed in the human mind or with simple assistance from conventional tools. The recitation of AI agents, embeddings, databases, and LLMs merely applies these mental processes using generic computers technology and does not remove the claim from the mental process category of abstract ideas.
Applicant further argues that the claimed use of embeddings, retrieval of query relevant content, and LLM processing improves computer functionality by reducing computational overhead and improving response generation. However, the Examiner maintains that the claim does not recite any specific technological improvement to embedding generation, retrieval techniques, database operation, AI model architecture, or LLM functionality; rather, the claim invokes the components at a high level of generality to perform their ordinary functions of processing, retrieving, and generating information. the alleged benefits identified by Applicant, such as improved relevance, reduced computational overhead, reduced post-processing, and generation of more helpful responses, are merely intended results and are not tied to any specific technical mechanism recited in the claim. Accordingly, the additional elements do not integrate the recited abstract idea into a practical application, and the claim remains directed to patent ineligible subject matter under step 2A, Prong Two.
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 1-7, 9-14, and 17-19 are rejected under 35 U.S.C. 101 because the claimed invention is directed to a judicial exception (i.e., a law of nature, a natural phenomenon, or an abstract idea) without significantly more.
The claim(s) recite(s) generating a response to the query by selecting a specific AI agent.
With respect to step 1 of the patent subject matter eligibility analysis, the claims are directed to a process, machine, manufacture, or composition of matter.
Independent claim 1 is directed to a method, which is a process.
Independent claim 9 is directed to a system which includes and processor and non-transitory computer readable media, which is directed to one of the four statutory subject matters.
Independent claim 16 directed to non-transitory computer readable media, which is directed to one of the four statutory subject matters.
Independent All other claims depend on claims 1, 9 and 16. As such, claims 1-20 are directed to a statutory category.
Regarding claims 1, 9 and 16:
With respect to step 2A, prong one (Judicial Exception), the claims recite an abstract idea, law of nature, or natural phenomenon. Specifically, the following limitations recite mathematical concepts and/or mental processes and/or certain methods of organizing human activity.
The claim recites the following limitations directed to an abstract idea:
Receiving a selection of an AI agent;
Receiving a query from a client device;
Determining whether the query is on-topic or off-topic;
Reformulate if off-topic;
Generating an embedding based on the query;
Retrieving query-relevant content from a knowledge database;
Determining a communication type;
Generating an LLM prompt using the query-relevant content;
Submitting a prompt to an LLM service and receiving an LLM output;
Returning a response.
These steps fall into recognized abstract idea:
Mathematical concept/algorithm (generating embeddings)
Mental process/organizing and/or analyzing information (retrieving content, determining communication type, constructing prompts);
Information management (selecting agents, routing queries, outputting response);
Using an LLM or AI model is merely considered as applying algorithm to data unless tied to a specific technological improvement.
There are no steps performed that provides a technical improvement to the computing system itself (improved data structure, improved model architecture, improved hardware).
Thus, the claims recite a judicial exception.
With respect to step 2A, Prong Two (Particular Application), the claims do not recite additional elements that integrate the judicial exception into a practical application. The following limitations are considered “additional elements” and explanation will be given as to why these “additional elements” do not integrate the judicial exception into a practical application.
The claims recite the use of:
Generic computing components (client device, knowledge database, LLM service),
Generic data manipulation (building prompts, retrieving content, returning a response),
Selecting among AI Agents.
These components merely use conventional computer components as tools to execute the abstract idea.
The limitations fail to transform the exception into a practical application. There is also no improvements to computer functionality or any specific technical solution to a computer centric problems (no improvements to how computers store data, or generate embeddings, run LLMs, communication or processing information). The claims focus on optimizing AI agents in a query and communication processing system, which are abstract improvements to information presentation and not technical improvements.
There is no recitation of, a new data structure that changes computer operation, improved communication, an unconventional indexing technique, a specific hardware solution.
Instead the claims recite conventional and generic computer functions performed in a routine manner, which does not amount to a practical application.
With respect to Step 2B. The claims do not include additional elements that are sufficient to amount to significantly more than the judicial exception. The recited components are merely generic computer/database elements performing their routine, well-understood, and conventional functions. See Alive, MPEP 2016.05(d).
The steps mentioned in the independent claims merely constitutes standard distributed-database behavior, such and basic replication, mirroring, and ownership transfer. Courts have consistently helped such high level information management operations are conventional.
The claims recite only functional, result oriented language (generating embeddings, retrieving contents via embedding, determining communication types, preparing LLM prompts, invoking LLM service), without specifying any technical mechanism for performing these operations in a non-conventional manner, and no new algorithmic mechanism and no new specific technical solution to a computer centric problem. All are routine, conventional operations of modern AI systems and digital information platforms.
Considering claims as a whole, the ordered combination of elements also reflects nothing more than the typical workflow of distributed systems, and therefore DOES NOT add “significantly more” than the abstract idea.
Such generic, high‐level, and nominal involvement of a computer or computer‐based elements for carrying out the invention merely serves to tie the abstract idea to a particular technological environment, which is not enough to render the claims patent‐eligible, as noted at pg.74624 of Federal Register/Vol. 79, No. 241, citing Alice, which in turn cites Mayo. Further, See, e.g., Alice Corp. Pty. Ltd. v. CLS Bank Int'l, 134 S. Ct. 2347, 2359‐60, 110 USPQ2d 1976, 1984 (2014). See also OIP Techs. v. Amazon.com, 788 F.3d 1359, 1364, 115 USPQ2d 1090, 1093‐94 (Fed. Cir. 2015) ("Just as Diehr could not save the claims in Alice, which were directed to 'implement[ing] the abstract idea of intermediated settlement on a generic computer', it cannot save O/P's claims directed to implementing the abstract idea of price optimization on a generic computer.") (citations omitted). See also, Affinity Labs of Texas LLC v. DirecTV LLC, 838 F.3d 1253, 1257‐1258 (Fed. Cir. 2016) (mere recitation of a GUI does not make a claimpatent‐eligible); Intellectual Ventures I LLC v. Capital One Bank, 792 F.3d 1363, 1370 (Fed. Cir. 2015) ("the interactive interface limitation is a generic computer element".).
The additional elements are broadly applied to the abstract idea at a high level of generality ("similar to how the recitation of the computer in the claims in Alice amounted to mere instructions to apply the abstract idea of intermediated settlement on a generic computer,") as explained in MPEP § 2106.05(f)) and they operate in a well‐understood, routine, and conventional manner.
MPEP § 2106.0S(d)(II) sets forth the following:
The courts have recognized the following computer functions as well-understood, routine, and conventional functions when they are claimed in a merely generic manner (e.g., at a high level of generality) or as insignificant extra-solution activity.
• Receiving or transmitting data over a network, e.g., using the Internet to gather data, Symantec ... ; TLI Communications LLC v. AV Auto. LLC ... ; OIP Techs., Inc., v. Amazon.com, Inc ... ; buySAFE, Inc. v. Google, Inc ... ;
• Performing repetitive calculations, Flook ... ; Bancorp Services v. Sun Life ... ;
• Electronic recordkeeping, Alice Corp ... ; Ultramercial ... ;
• Storing and retrieving information in memory, Versata Dev. Group, Inc. v. SAP Am., Inc ... ;
• Electronically scanning or extracting data from a physical document, Content Extraction and Transmission, LLC v. Wells Fargo Bank ... ; and
• A web browser's back and forward button functionality, Internet Patent
• Corp. v. Active Network, Inc. ...
. . . Courts have held computer-implemented processes not to be significantly more than an abstract idea (and thus ineligible) where the claim as a whole amounts to nothing more than generic computer functions merely used to implement an abstract idea, such as an idea that could be done by a human analog (i.e., by hand or by merely thinking).
In addition, when taken as an ordered combination, the ordered combination adds nothing that is not already present as when the elements are taken individually. There is no indication that the combination of elements integrate the abstract idea into a practical application. Their collective functions merely provide conventional computer implementation. Therefore, when viewed as a whole, these additional claim elements do not provide meaningful limitations to transform the abstract idea into a practical application of the abstract idea or that the ordered combination amounts to significantly more than the abstract idea itself.
The dependent claims have been fully considered as well, however, similar to the findings for claims above, these claims are similarly directed to the “Mental Processes” grouping of abstract ideas set forth in the 2019 PEG, without integrating it into a practical application and with, at most, a general purpose computer that serves to tie the idea to a particular technological environment, which does not add significantly more to the claims. The ordered combination of elements in the dependent claims (including the limitations inherited from the parent claim(s)) add nothing that is not already present as when the elements are taken individually. There is no indication that the combination of elements improves the functioning of a computer or improves any other technology. Their collective functions merely provide conventional computer implementation. Accordingly, the subject matter encompassed by the dependent claims fails to amount to significantly more than the abstract idea.
Looking at the claim as a whole does not change this conclusion and the claim is ineligible.
Regarding claims 2, and 10,
The claim recites:
Selecting AI agent based on domain, intent/purpose, user preferences, user characteristics, or system constraints.
This is merely an information analysis: determining domain/intent/preferences (mental process). Selecting an AI agent based on such criteria is organizing and classifying information, not a technical improvement.
This does not change the nature of the abstract idea. It does not add a technical improvement to an abstract idea, such as improving computer functionality, data structure, or processing architecture.
There is no practical application, and no inventive step, the claims are still considered abstract.
Regarding claims 3, and 11,
The claim recites:
Selecting AI agent based on user input via GUI, selection criteria, user profile, or interaction history.
This is merely data evaluation: receiving and applying user input or historical interaction data. GUI selection is generic user interface activity, not a technical improvement. Users profiles and history are informational and not tied to any improved database or computing mechanism.
This does not change the nature of the abstract idea. It does not add a technical improvement to an abstract idea, such as improving computer functionality, data structure, or processing architecture.
There is no practical application, and no inventive step, the claims are still considered abstract.
Regarding claim 4,
The claim recites:
Retrieving conversation history, augmenting the query with relevant portions, generating embedding from augmented query.
This is merely information retrieval and manipulation: retrieving and using conversation history. Augmenting a query is linguistic/mathematical operation (mathematical concept). There is no improvement to embedding generation memory, or model architecture.
This does not change the nature of the abstract idea. It does not add a technical improvement to an abstract idea, such as improving computer functionality, data structure, or processing architecture.
There is no practical application, and no inventive step, the claims are still considered abstract.
Regarding claims 5, 12 and 17,
The claim recites:
Query rewriting to correct spelling/grammar, clarify terms, remove irrelevant information and simplify questions before embedding.
This is merely content modification based on rules/policies: text cleaning, rewriting and simplifying. Augmenting a query is linguistic/mathematical operation (mathematical concept). This does not change how the embeddings are generated, and only modifies input data. No improvements to computer performance or efficiency.
This does not change the nature of the abstract idea. It does not add a technical improvement to an abstract idea, such as improving computer functionality, data structure, or processing architecture.
There is no practical application, and no inventive step, the claims are still considered abstract.
Regarding claims 6, 13 and 18,
The claim recites:
Determining communication tone/style (formal, humorous, concise, direct, persuasive, etc.).
This is merely semantic classification: tone/style selection (mental process). This does not improve any technical systems and only modifies the structure of the response content. No improvements to computer performance beyond generic application of rules.
This does not change the nature of the abstract idea. It does not add a technical improvement to an abstract idea, such as improving computer functionality, data structure, or processing architecture.
There is no practical application, and no inventive step, the claims are still considered abstract.
Regarding claims 7, 14 and 19,
The claim recites:
Identifying communication attributes, generating instruction based on the attributes, and constructing an LLM prompt.
This is merely information evaluation: identifying communication attributes (mental process). Data manipulation does not improve any technical systems and generates instruction and formats the prompts. No improvements or new prompt architecture, model training method; or computational improvements.
This does not change the nature of the abstract idea. It does not add a technical improvement to an abstract idea, such as improving computer functionality, data structure, or processing architecture.
There is no practical application, and no inventive step, the claims are still considered abstract.
Regarding claims 8, 15 and 20,
The claim recites:
Classifying a query as an on/off topic; reformulating the off-topic queries to make them on-topic; generating a new embedding.
This is merely a classification algorithm: topic classification (mathematical/mental process). Reformulating queries is semantic rewriting, which is an abstract operation. No improvements to LLM performance, memory structure, or computational technique.
This does not change the nature of the abstract idea. It does not add a technical improvement to an abstract idea, such as improving computer functionality, data structure, or processing architecture.
There is no practical application, and no inventive step, the claims are still considered abstract.
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.
Claim(s) 1-3, 6, 7, 9-11, 13, 14, 16, 18, and 19 are rejected under 35 U.S.C. 103 as being unpatentable over Crabtree; Jason et al. (US 20250258685 A1) [Crabtree-1] in view of Gomez; Laurent (US 20250165648 A1) [Gomez].
Regarding claims 1, 9 and 16, Crabtree-1 discloses, a computer-implemented method comprising: selecting a particular generative artificial intelligence (AI) agent, from among a set of one or more AI agents, the particular generative artificial intelligence (AI) agent associated with an entity (Agent orchestration system 2232 may perform prompt engineering tasks to create one or more prompts based on the user specification and the selected Gen AI systems to be submitted to the selected Gen AI systems. The selected agents may then generate UX/UI content based on the prompt ¶ [0224]. FIG. 24 is a block diagram illustrating an exemplary aspect of dynamic application experience generation platform, an agent orchestration system 2400. According to the aspect, agent orchestration system 2400 comprises an agent selector subsystem 2401, a prompt engineering subsystem 2402, and one or more agents 2403a-n which represent one or more Gen AI systems. According to the aspect, agent orchestration system 2400 receives a user specification from design management system 2400 via agent selector 2401. Agent selector 2401 may be configured to parse the user specification and select one or more appropriate Gen AI systems (also referred to herein as agents) to generate the UX/UI content described by the user specification ¶ [0234] and [0236]-[0237]);
receiving a use query, the query sent by a client device (to received user queries ¶ [0097]. According to the embodiment, embedding model 615 may also receive a user query from experience curation 640 and vectorize it where it may be stored in vector database 620. This provides another useful datapoint to provide deeper context when comparing received queries against stored query embeddings ¶ [0108]-[0109], [0122], [0116]);
generating an embedding based on the reformulated query (Embedding refinement 1922 may utilize algorithms that can query the knowledge graph to retrieve relevant facts, entities, and relationships based on the input data and the current reasoning context. Retrieved knowledge may be used to refine the vector embeddings generated by the neural networks models, incorporating symbolic information into the distributed representations (embeddings). In some implementations, techniques like attention mechanisms, graph convolutions, and/or knowledge-aware language models to effectively combine the embeddings with the knowledge graph data. For example, consider a recommendation system that generates initial embeddings for users and items based on their interaction history. The platform then queries a knowledge graph of user demographics, item categories, and contextual factors (e.g., time, location) to retrieve relevant information ¶ [0151]. According to the embodiment, embedding model 615 may also receive a user query from experience curation 640 and vectorize it where it may be stored in vector database 620. This provides another useful datapoint to provide deeper context when comparing received queries against stored query embeddings ¶ [0108]-[0109], [0122]. Prompt engineering is an iterative process that involves designing and refining prompts to help Gen AI systems perform specific tasks (e.g., UX/UI content generation) effectively ¶ [0236], [0249]);
using the embedding to retrieve query-relevant content associated with the entity from a knowledge database that stores content associated with the entity (Embeddings are dense vector representations that capture the semantic meaning and relationships of data points. Vector databases 1928 store and index these embeddings for efficient retrieval and similarity search.… The refined embeddings are stored in a vector database for fast retrieval during recommendation generation ¶ [0146]. Embedding refinement 1922 may utilize algorithms that can query the knowledge graph to retrieve relevant facts, entities, and relationships based on the input data and the current reasoning context. Retrieved knowledge may be used to refine the vector embeddings generated by the neural networks models, incorporating symbolic information into the distributed representations (embeddings). In some implementations, techniques like attention mechanisms, graph convolutions, and/or knowledge-aware language models to effectively combine the embeddings with the knowledge graph data ¶ [0151], [0162]);
determining a communication type associated with the entity (For example, a response may first be filtered of any personal information by curation 713a prior to the being relayed back to the user. As another example, curation 713a may transform the response into specific format, style, or language based on user defined preferences ¶ [0115]. This information ensures that the chatbot adheres to the user's privacy requirements and aligns with their preferred communication style ¶ [0332]. Examiner specifies that response with a specific format, style, or language has been interpreted as communication type);
returning a response to the query based on the particular LLM output (In turn, the LLM processes the prompts, contextual data, and user query to generate a contextually aware response which can be sent to experience curation 640 where the response may be curated, or not, and returned to the user as output 604 ¶ [0110]).
However, Crabtree-1 does not explicitly facilitate generating a large language model (LLM) prompt comprising the query relevant content and instructions to apply the communication type to LLM output; receiving a particular LLM output from a LLM service in response to submitting the LLM prompt to the LLM service.
Gomez discloses, generating a large language model (LLM) prompt comprising the query relevant content and instructions to apply the communication type to LLM output (In any of the examples herein, prompts can be provided to LLMs to generate responses. Prompts in LLMs can be initial input instructions that guide model behavior. Prompts can be textual cues, questions, or statements that users provide to elicit desired responses from the LLMs. Prompts can act as primers for the model's generative process. Sources of prompts can include user-generated queries, predefined templates, or system-generated suggestions. Technically, prompts are tokenized and embedded into the model's input sequence, serving as conditioning signals for subsequent text generation. Users can experiment with prompt variations to manipulate output, using techniques like prefixing, temperature control, top-K sampling, etc. These prompts, sourced from diverse inputs and tailored strategies, enable users to influence LLM-generated content by shaping the underlying context and guiding the neural network's language generation. For example, prompts can include instructions and/or examples to encourage the LLMs to provide results in a desired style and/or format ¶ [0032]); receiving a particular LLM output from a LLM service in response to submitting the LLM prompt to the LLM service (prompts can include instructions and/or examples to encourage the LLMs to provide results in a desired style and/or format ¶ [0032]. Also see ¶ [0074], [0102], [0131]).
It would have been obvious to one ordinary skilled in the art before the effective filing date of the claimed invention to combine the teachings of the cited references because Gomez’s system would have allowed Crabtree-1 to facilitate generating a large language model (LLM) prompt comprising the query relevant content and instructions to apply the communication type to LLM output; receiving a particular LLM output from a LLM service in response to submitting the LLM prompt to the LLM service. The motivation to combine is apparent in the Crabtree-1’s reference, because there is a need for improving data security when integrating generative AI into an enterprise environment.
However, neither Crabtree-1 nor Gomez explicitly facilitate analyzing the user query using an on-topic classifier module to determine whether the user query is on-topic or off-topic for the particular generative Al agent; if the user query is determined to be off-topic: identifying a key entity from the user query that is relevant to the particular generative Al agent's knowledge domain; and reformulating the user query to be on-topic to the particular generative Al agent's knowledge domain by modifying the query based on the identified key entity.
Mehrotra discloses, analyzing the user query using an on-topic classifier module to determine whether the user query is on-topic or off-topic for the particular generative Al agent (evaluator, assess whether a prompt (generated by A) is off-topic for G (i.e., evaluate the Off-Topic function) ¶ [0045], [0149], pruning all off-topic prompts before querying/prompt is refined ¶ [0043], [0048], [0051], [0055], [0057]-[0059], [0071]-[0074], [0152]-[0154], [0157], [0158], retaining on-topic and pruning off-topic ¶ [0164]).
if the user query is determined to be off-topic (a function such that Off-Topic(P, G) is 1 if P is off-topic for G and it is 0 otherwise ¶ [0036], [0043], evaluate the Off-Topic function ¶ [0045], If Off-Topic(P,G) = True, then delete [page 4, Algorithm 1: Branch-and-Prune table], if Off-Topic(P, G)=1, then the node corresponding to prompt P is pruned [0154]):
reformulating the user query to be on-topic to the particular generative Al agent's knowledge domain by modifying the query based on the identified key entity (iteratively refine G ¶ [0045], prompt P is refined… refinement iteration… generates the improved prompt ¶ [0153], the Prompt Automatic Iterative Refinement (PAIR) … to revise the prompt ¶ [0158]. pruning all off-topic prompts before querying/prompt is refined ¶ [0043], [0048], [0051], [0055], [0057]-[0059], [0071]-[0074], [0157], retaining on-topic and pruning off-topic ¶ [0164]).
It would have been obvious to one ordinary skilled in the art before the effective filing date of the claimed invention to combine the teachings of the cited references because Mehrotra’s system would have allowed Crabtree-1 and Gomez to facilitate analyzing the user query using an on-topic classifier module to determine whether the user query is on-topic or off-topic for the particular generative Al agent; if the user query is determined to be off-topic: and reformulating the user query to be on-topic to the particular generative Al agent's knowledge domain by modifying the query based on the identified key entity. The motivation to combine is apparent in the Crabtree-1 and Gomez’s reference, because there is a need for improving prompt generation using jailbreak black-box large language models.
However, neither one of Crabtree-1, Gomez or Mehrotra explicitly facilitates identifying a key entity from the user query that is relevant to the particular generative Al agent's knowledge domain.
Dar discloses, identifying a key entity from the user query that is relevant to the particular generative Al agent's knowledge domain (topic extraction is a natural language programming technique that automatically identifies the main key phrases, topics, or themes… example, suppose query is “How do I determine the placement of the new expansion enclosure?” … generation process extracts a topic (e.g., query topic ) concerning “new expansion enclosure”…extracting the topic includes extracting a plurality of topics for the query … query representation generation process further extracts “placement” as query topic and “determine” as query topic ¶ [0062]-[0063]).
It would have been obvious to one ordinary skilled in the art before the effective filing date of the claimed invention to combine the teachings of the cited references because Dar’s system would have allowed Crabtree-1, Gomez and Mehrotra to facilitate identifying a key entity from the user query that is relevant to the particular generative Al agent's knowledge domain. The motivation to combine is apparent in the Crabtree-1, Gomez and Mehrotra’s reference, because there is a need for improving information retrieval stage, namely effective chunking, chunk indexing, and the similarity search for a small set of chunks (out of thousands and potentially millions) that are the best match for a given query may have a much greater impact on query processing performance than the final LLM stage.
Regarding claims 2 and 10, the combination of Crabtree-1, Gomez, Mehrotra and Dar discloses, wherein the selection of the particular generative AI agent is based on one or more of: (a) analyzing the user query to determine a domain or subject matter associated with the user query, and selecting the particular generative AI agent based on the domain or subject matter; (b) analyzing the user query to determine an intent or purpose associated with the user query, and selecting the particular generative AI agent based on the intent or purpose; (c) analyzing user data associated with the client device to determine user preferences or characteristics, and selecting the particular generative AI agent based on the user preferences or characteristics; or (d) analyzing system data associated with the multi-user application system to determine system constraints or requirements, and selecting the particular generative AI agent based on the system constraints or requirements (Crabtree-1: Agent selector 2401 may be configured to parse the user specification and select one or more appropriate Gen AI systems (also referred to herein as agents) to generate the UX/UI content described by the user specification. The selection of the one or more agents may be based on various factors including, but not limited to, the user defined requirements (e.g., target audience, design goals, functionality, platform/device, etc.), Gen AI (gen AI) system compatibility (e.g., using an LLM to generate text, diffusion models to generate images or sound, etc.), model performance (e.g., factors such as the quality of designs, the range of design options, and the ability to customize to meet the user's needs), model integration (e.g., models which can easily be integrated into existing workflows and tools), cost and licensing, and user/expert feedback (e.g., gathered feedback from stakeholders and iterate on design) ¶ [0234]. The AI agent could then use this information to dynamically modify or navigate the site in a way that is organic and tailored to this particular user's needs and preferences for engagement (e.g., browser vs. a search vs. an explorer etc.) ¶ [0241]. Also see ¶ [0167]).
Regarding claims 3 and 11, the combination of Crabtree-1, Gomez, Mehrotra and Dar discloses, wherein the selection of the particular generative AI agent is made by a user of the client device based on one or more of: (a) receiving a user input specifying the particular generative AI agent from a list of available generative AI agents presented to the user in a graphical user interface (GUI); (b) receiving a user input specifying criteria for selecting the particular generative AI agent, such as a desired domain, subject matter, or communication style, and a subsequent selection of the particular generative AI agent based on the user-specified criteria; (c) accessing, a user profile associated with the user, the user profile indicating a preferred generative AI agent or preferences for selecting generative AI agents; or (d) processing a user interaction history associated with the user, the user interaction history indicating previous selections or preferences for generative AI agents by the user (Crabtree-1: According to the aspect, agent orchestration system 2400 receives a user specification from design management system 2400 via agent selector 2401. Agent selector 2401 may be configured to parse the user specification and select one or more appropriate Gen AI systems (also referred to herein as agents) to generate the UX/UI content described by the user specification. The selection of the one or more agents may be based on various factors including, but not limited to, the user defined requirements (e.g., target audience, design goals, functionality, platform/device, etc.), Gen AI (gen AI) system compatibility (e.g., using an LLM to generate text, diffusion models to generate images or sound, etc.), model performance (e.g., factors such as the quality of designs, the range of design options, and the ability to customize to meet the user's needs), model integration (e.g., models which can easily be integrated into existing workflows and tools), cost and licensing, and user/expert feedback (e.g., gathered feedback from stakeholders and iterate on design) ¶ [0234], [0237], [0239]).
Regarding claims 6, 13 and 18, the combination of Crabtree-1, Gomez, Mehrotra and Dar discloses, wherein the communication type associated with the entity indicates a conversational tone or style of the entity, and wherein the conversational tone or style comprises one or more of: a formal or informal tone; a friendly or professional tone; a humorous or serious tone; a concise or elaborate style; a direct or indirect style; an empathetic or neutral tone; or a persuasive or informative tone (Crabtree-1: The rendered content can prioritize displaying content from these sources. Users may prefer content with a certain tone or style, such as formal, casual, humorous, or informative. The rendered content can adjust the tone and style of content displayed based on the user's preference. Users may prefer to see content that is updated frequently or prefer a more static content experience. The platform can adjust the frequency of content updates based on the user's preference ¶ [0280]. The prompt should be clear, concise, and include any relevant information or examples that the model needs to generate a response. In some implementations, system 2300 may experiment with different prompts and parameters to see how they affect the model's performance. This may involve adjusting the length of the prompt, the type of information included, and other factors ¶ [0236]. Additionally, user preferences regarding the chatbot's behavior, tone, and interaction style are collected. This information ensures that the chatbot adheres to the user's privacy requirements and aligns with their preferred communication style ¶ [0332]).
Regarding claims 7, 14 and 19, the combination of Crabtree-1, Gomez, Mehrotra and Dar discloses, wherein generating the LLM prompt comprises: analyzing the communication type associated with the entity to identify one or more communication attributes, wherein the communication attributes specify characteristics of the entity's communication style (Crabtree-1: Additionally, user preferences regarding the chatbot's behavior, tone, and interaction style are collected. This information ensures that the chatbot adheres to the user's privacy requirements and aligns with their preferred communication style ¶ [0332]. Also see ¶ [0236] and [0280]);
generating instructions based on the identified communication attributes, wherein the instructions guide the LLM service to generate the particular LLM output in accordance with the entity's communication style (Crabtree-1: Additionally, user preferences regarding the chatbot's behavior, tone, and interaction style are collected. This information ensures that the chatbot adheres to the user's privacy requirements and aligns with their preferred communication style ¶ [0332]. Also see ¶ [0236] and [0280]. LLMs have become increasingly popular because they have broad applicability for a range of NLP tasks, including but not limited to, text generation, translation, content summary, rewriting content, classification and categorization, sentiment analysis, and conversational AI and chatbots ¶ [0179]); and
constructing the LLM prompt to comprise: the reformulated query or a rewritten or augmented version of the reformulated query (Gomez: receiving a prompt query entered through a user interface, detecting sensitive data in the prompt query that violates a security protocol, generating a modified prompt query which anonymizes the sensitive data, submitting the modified prompt query to a large language model, and receiving a reply generated by the large language model [Abstract]. Also see ¶ [0172]-[0173]);
the query-relevant content retrieved from the knowledge database (Crabtree-1: According to the embodiment, embedding refinement computing system 1922 is present and configured to incorporate data from one or more knowledge graphs. Embedding refinement 1922 may utilize algorithms that can query the knowledge graph to retrieve relevant facts, entities, and relationships based on the input data and the current reasoning context ¶ [0151], [0162]); and
the generated instructions for applying the communication type to LLM output to generate the particular LLM output in accordance with the entity’s communication style (Crabtree-1: Additionally, user preferences regarding the chatbot's behavior, tone, and interaction style are collected. This information ensures that the chatbot adheres to the user's privacy requirements and aligns with their preferred communication style ¶ [0332]. Also see ¶ [0236] and [0280]).
Claim(s) 4 are rejected under 35 U.S.C. 103 as being unpatentable over Crabtree-1 in view of Gomez in view of Bista; Umanga et al. (US 20250094455 A1) [Bista].
Regarding claim 4, the combination of Crabtree-1, Gomez, Mehrotra and Dar teaches all the limitation of claim 1.
However, neither one of Crabtree-1, Gomez, Mehrotra or Dar explicitly facilitates wherein generating the embedding based on the reformulated query comprises: retrieving a conversation history associated with the client device or the user of the client device, the conversation history comprising one or more previous queries and corresponding responses between the user and the generative AI agent; augmenting the reformulated query with the retrieved conversation history to generate an augmented query, wherein the augmented query includes the reformulated query and relevant portions of the conversation history; and generating the embedding based on the augmented query using the embedding generator, wherein the embedding represents a semantic understanding of the reformulated query in the context of the conversation history.
Bista discloses, wherein generating the embedding based on the reformulated query in comprises: retrieving a conversation history associated with the client device or the user of the client device, the conversation history comprising one or more previous queries and corresponding responses between the user and the generative AI agent (CQR stands for contextual query rewriting. The purpose of CQR is to ensure that every user utterance contains all relevant context from the previous conversation history. The system contextualizes user messages because the pipeline does not ingest or interpret the context from conversation history on its own. Thus, CQR generates a modified query that contains the appropriate context for the pipeline to retrieve the right documents and generate an answer to the user's query ¶ [0027]. In some embodiments, the CQR model 214C is applied to rewrite a query based on conversational history, significantly enhance the interpretation of the query by the response LLM 216B. In some aspects, the query rewrite includes two components. First, context understanding is performed. Conversation history is used to rewrite the current user query (e.g., contextualize anaphoras and ellipses using the history) so that the query can be independently answered by the response LLM 216B ¶ [0078]);
augmenting the reformulated query with the retrieved conversation history to generate an augmented query, wherein the augmented query includes the reformulated query and relevant portions of the conversation history; and generating the embedding based on the augmented query using the embedding generator, wherein the embedding represents a semantic understanding of the reformulated query in the context of the conversation history (At step 422, the C2K frontend 404 provides the query q.sub.n to the CQR model 408. The CQR model 408 rewrites the query using conversation history 406. The conversation history 406 includes queries and corresponding answers (q.sub.1, a.sub.1), (q.sub.2, a.sub.2), . . . (q.sub.n−1, a.sub.n−1), where an is the answer to the nth query. The rewritten query q*.sub.n is provided to the C2K frontend at step 424. In some examples, the CQR model 408 uses the previous w question-answer pairs to resolve ambiguities in the query q.sub.n, where w is a configurable number (e.g., w=3 is suitable). The query rewriting process is further described below with respect to FIG. 5 ¶ [0083]. Also see [abstract], ¶ [0006] and [0078]).
It would have been obvious to one ordinary skilled in the art before the effective filing date of the claimed invention to combine the teachings of the cited references because Bista’s system would have allowed Crabtree-1, Gomez, Mehrotra and Dar to facilitate wherein generating the embedding based on the reformulated query comprises: retrieving a conversation history associated with the client device or the user of the client device, the conversation history comprising one or more previous queries and corresponding responses between the user and the generative AI agent; augmenting the reformulated query with the retrieved conversation history to generate an augmented query, wherein the augmented query includes the reformulated query and relevant portions of the conversation history; and generating the embedding based on the augmented query using the embedding generator, wherein the embedding represents a semantic understanding of the reformulated query in the context of the conversation history. The motivation to combine is apparent in the Crabtree-1, Gomez, Mehrotra and Dar’s reference, because there is a need to improve contextualizing a query to be processed by a language model.
Claim(s) 5, 8, 12, 15, 17 and 20 are rejected under 35 U.S.C. 103 as being unpatentable over Crabtree-1 in view of Gomez in view of Crabtree; Jason et al. (US 20250259085 A1) [Crabtree-2].
Regarding claims 5, 12 and 17, the combination of Crabtree-1, Gomez, Mehrotra and Dar teaches all the limitations of claim 1.
However, neither one of Crabtree, Gomez, Mehrotra or Dar explicitly facilitates wherein generating the embedding based on the reformulated query comprises: analyzing the reformulated query using a query rewriting module to identify one or more of: spelling or grammatical errors in the reformulated query; ambiguous or unclear terms in the reformulated query; irrelevant or unnecessary information in the reformulated query; or complex or compound questions in the reformulated query; rewriting the reformulated query based on the analysis to generate a rewritten query, wherein the rewritten query corrects spelling or grammatical errors, clarifies ambiguous terms, removes irrelevant information, or simplifies complex questions; and generating the embedding based on the rewritten query using the embedding generator, wherein the embedding represents a semantic understanding of the rewritten query.
Crabtree-2 discloses, wherein generating the embedding based on the reformulated query in comprises: analyzing the reformulated query using a query rewriting module to identify one or more of: spelling or grammatical errors in the reformulated query; ambiguous or unclear terms in the reformulated query; irrelevant or unnecessary information in the reformulated query; or complex or compound questions in the reformulated query; rewriting the reformulated query based on the analysis to generate a rewritten query, wherein the rewritten query corrects spelling or grammatical errors, clarifies ambiguous terms, removes irrelevant information, or simplifies complex questions; and generating the embedding based on the rewritten query using the embedding generator, wherein the embedding represents a semantic understanding of the rewritten query (A memory query augmentation system (MQAS) 6080 employs large language models to dynamically reformulate ambiguous or contextually imprecise queries, enhancing retrieval accuracy. Through an iterative reinforcement learning mechanism that leverages downstream task-specific performance metrics, MQAS 6080 continuously refines its query augmentation and retrieval strategy. This adaptive, reinforcement-driven augmentation method significantly elevates the accuracy, relevance, and consistency of memory recall across various modalities including episodic, semantic, and procedural memory systems [0324]. Also see ¶ [0314], [0322]).
It would have been obvious to one ordinary skilled in the art before the effective filing date of the claimed invention to combine the teachings of the cited references because Crabtree-2’s system would have allowed Crabtree-1, Gomez, Mehrotra and Dar to facilitate wherein generating the embedding based on the reformulated query comprises: analyzing the reformulated query using a query rewriting module to identify one or more of: spelling or grammatical errors in the reformulated query; ambiguous or unclear terms in the reformulated query; irrelevant or unnecessary information in the reformulated query; or complex or compound questions in the reformulated query; rewriting the reformulated query based on the analysis to generate a rewritten query, wherein the rewritten query corrects spelling or grammatical errors, clarifies ambiguous terms, removes irrelevant information, or simplifies complex questions; and generating the embedding based on the rewritten query using the embedding generator, wherein the embedding represents a semantic understanding of the rewritten query. The motivation to combine is apparent in the Crabtree-1, Gomez, Mehrotra and Dar’s reference, because there is a need to improve scalable platforms that enable secure and optionally privacy-aware knowledge exchange and negotiation between domain-specialized artificial intelligence agents through modular hybrid computing architectures.
Regarding claims 8, 15 and 20, the combination of Crabtree-1, Gomez, Mehrotra, Dar and Crabtree-2 discloses, prior to generating the embedding based on the query: analyzing the query using an on-topic classifier module to determine whether the query is on-topic or off-topic for the particular generative AI agent; if the query is determined to be off-topic: identifying one or more key entities, concepts, or themes from the query that are relevant to the generative AI agent's knowledge domain; reformulating the query by modifying, expanding, or narrowing its scope based on the identified key entities, concepts, or themes to generate a reformulated query that is on-topic for the generative AI agent; replacing the original query with the reformulated query for subsequent processing steps; and generating the embedding based on the reformulated on-topic query, wherein the embedding represents a semantic understanding of the reformulated query (Crabtree-2: A memory query augmentation system (MQAS) 6080 employs large language models to dynamically reformulate ambiguous or contextually imprecise queries, enhancing retrieval accuracy. Through an iterative reinforcement learning mechanism that leverages downstream task-specific performance metrics, MQAS 6080 continuously refines its query augmentation and retrieval strategy. This adaptive, reinforcement-driven augmentation method significantly elevates the accuracy, relevance, and consistency of memory recall across various modalities including episodic, semantic, and procedural memory systems ¶ [0324]. Also see ¶ [0322]-[0323]. Examiner has interpreted “reformulate ambiguous or contextually imprecise queries” as reformulating the off-topic query; “dynamically reformulate” has been interpreted as replacing the original query; “memory query augmentation” is inherently prior to embedding).
Conclusion
THIS ACTION IS MADE FINAL. 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 MOHAMMAD S ROSTAMI whose telephone number is (571)270-1980. The examiner can normally be reached Mon-Fri From 9 a.m. to 5 p.m..
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, Boris Gorney can be reached at (571)270-5626. 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.
6/3/2026
/MOHAMMAD S ROSTAMI/Primary Examiner, Art Unit 2154
Prosecution Insights