DETAILED ACTION
This communication is in response to the Application filed on 06/04/2024 (provisional). Claims 1-20 are pending and have been examined.
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 .
Examiner Note for Double Patenting
Although no double patenting was issued, the examiner has the right to include one in the future based on change of scope. Claims 1-20 have been examined for double patenting in view of the following references:
U.S. Patent Application No. 18/911,868, “USER-SPECIFIC SYSTEM PROMPTS FOR LANGUAGE GENERATION”
U.S. Patent Application No. 18/911,903, “USER AND OPERATOR PREFERENCE RECONCILIATION FOR PRE-PROMPT ENGINEERING”
U.S. Patent Application No. 18/911,882, “USER- AND OPERATOR-SPECIFIC SYSTEM PROMPTS FOR LANGUAGE GENERATION”
Claim Objections
Claim 5 objected to because it mentions the at least one first operator preference being dependent on claim 1; however, the at least one first operator preference is first introduced in claim 3 and is not explicitly mentioned nor defined in claim 1:
“The method of claim 1, wherein receiving the at least one first operator preference comprises querying, by the server, a first database using the user identifier to retrieve the at least one first operator preference.” should read:
The method of claim 3, wherein receiving the at least one first operator preference comprises querying, by the server, a first database using the user identifier to retrieve the at least one first operator preference.”
Claim 6 objected to because it mentions receiving at least one second operator preference being dependent on claim 1; however, the at least one first operator preference is first introduced in claim 3 and further disclosed in claim 5. The at least first operator preference is not explicitly mentioned nor defined in claim 1:
“The method of claim 1, and further comprising receiving at least one second operator preference indicative of at least one third characteristic, preferred by the operator of the server, of the natural-language outputs, wherein modifying the system prompt comprises modifying the system prompt based on the on the retrieved at least one user preference, the at least one first operator preference, and the at least one second operator preference to generate the modified system prompt.” should read:
The method of claim 5, and further comprising receiving at least one second operator preference indicative of at least one third characteristic, preferred by the operator of the server, of the natural-language outputs, wherein modifying the system prompt comprises modifying the system prompt based on the on the retrieved at least one user preference, the at least one first operator preference, and the at least one second operator preference to generate the modified system prompt.
Claim 10 objected to because of the following informalities:
“The method of claim 1, and further comprising querying, by the processor, a third database with the user identifier to retrieve third information, and wherein generating, by the processor, the vector embedding comprises a generating a vector embedding representative of the first information, the third information, and the natural-language prompt.” should read:
The method of claim 1, and further comprising querying, by the processor, a third database with the user identifier to retrieve third information, and wherein generating, by the processor, the vector embedding comprises generating a vector embedding representative of the first information, the third information, and the natural-language prompt.
Claim 19 objected to because it mentions the at least one first operator preference being dependent on claim 1; however, the at least one first operator preference is first introduced in claim 3 and is not explicitly mentioned nor defined in claim 1:
“The method of claim 17, wherein the at least one first operator preference comprises a second natural-language word.” should read:
The method of claim 3, wherein the at least one first operator preference comprises a second natural-language word.
Appropriate correction is required.
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION. —The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claims 1-20 rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention.
Claim 1 and claim 20 recite the limitation “the server” in page 1, line 6, and page 6, line 2, respectively. There is insufficient antecedent basis for this limitation in the claim.
Claim 1 recites the limitation “the processor” in line 17, respectively. There is insufficient antecedent basis for this limitation in the claim.
Dependent claims are rejected for being dependent upon an indefinite base claim.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claim(s) 1, 2, 5, 10, 11, 12, 16, 17, 18, and 19 is/are rejected under 35 U.S.C. 103 as being unpatentable over Crabtree (US 20250258685 A1) in view of Mann (US 20250217576 A1)).
Regarding claim 1, Crabtree teaches the following limitations:
A method of natural language generation, the method comprising: receiving, by a user device, an indication of at least one user preference for a user, the at least one user preference indicative of at least one first characteristic, preferred by user, of natural-language outputs generated by a machine learning language model based on natural-language text inputs (see [0012], where a method for user experience curation, comprising the steps of: collecting a plurality of data types from a plurality of sources or systems wherein data may include but is not limited to audio, visual, telematics, haptic, smells, environmental conditions, user security settings, user privacy settings, user disabilities and sensory ranges, user health and faculties, other user devices, and user preference settings. Also see [0279], where preference database 2800 may store a plurality of user interaction preferences 2810);
receiving,
by the server (see [0093] FIG. 4 is a block diagram illustrating an exemplary system architecture for a distributed generative artificial intelligence reasoning and action platform 420, according to an embodiment. According to the embodiment, platform 420 is configured as a cloud-based computing platform comprising various system or sub-system components configured to provide functionality directed to the execution of neuro-symbolic Gen AI reasoning and action)
and from the user device, the at least one user preference (see [0012] where a method for user experience curation, comprising the steps of: collecting a plurality of data types from a plurality of sources or systems wherein data may include but is not limited to audio, visual, telematics, haptic, smells, environmental conditions, user security settings, user privacy settings, user disabilities and sensory ranges, user health and faculties, other user devices, and user preference settings. Also see [0264], where a context extraction and tracking system 2731 is present and configured to receive the content request from a user device 2711, format the request and submit it to the service provider/content provider on the user's behalf, receive un-curated content from the service provider/content provider, and extract and classify relevant content from the received un-curated content);
modifying, by the server, a system prompt for the machine-learning language model based on the received at least one user preference to generate a modified system prompt (see [0272], where user preference data is retrieved. A prompt is engineered using the outputs from the classification tasks performed on both devices and the user preference data);
providing, by the server, the modified system prompt as an initial input to the machine-learning language model (see [0272], where the prompt and content elements are submitted to Gen AI models. The output of the Gen AI models is rendered on the user interface of the second device as curated content responsive to the content request);
receiving, by a server and from the user device, a natural-language text prompt and a user identifier for the user, the natural-language text prompt provided by the user to a chat application operating on the user device (see [0109], where a user may submit a query 603 to an experience curation engine 640 which starts the prompt construction and retrieval process. The query is sent to DCG 630 which can send the query to various components such as prompt engineering 625 and embedding model 615. Embedding model 615 receives the query and vectorizes it and stores it in vector database 620. The vector database 620 can send contextual data (via vectors) to DCG 630 and to various APIs/plugins 635. Prompt engineering 625 can receive prompts 602 from developers to train the model on. Also see [0263], where dynamic experience curation platform 2730 comprises various components which support and provide capabilities directed to experience curation for platform users 2710. A user management system 2735 is present and configured to provide a portal for each user to create and manage their own user profile. A user profile may be created and stored in a preference database 2736. The user profile may comprise information associated with the user including, but not limited to, user contact information (e.g., name, email address, address, IP address, etc.), demographic information (e.g., age, ethnicity, political party, etc.), login information (e.g., username and password), and a plurality of preferences);
querying, by the server, a first database with the user identifier to retrieve first information (see [0276], where a preference database 2736 is present and configured to store a user profile and a plurality of user preferences which may be explicitly defined/declared by the user and/or implicitly learned/derived from user behavior and interactions. Preference database 2736 may also store rules or policies or other legal considerations which may be applicable to a user and/or requested content based on preferences, location, age, or any other constraint. Also see [0271], where a session management system 2733 may be configured to support a user's curated experience between and across multiple user devices 2711. Session management systems 2733 may manage a user session across multiple devices. In some implementations, this may comprise steps involving assigning a unique user identifier (ID) (e.g., username, email address, or generated ID, etc.) to identify the user consistently across devices and sessions. Also see [0263], where dynamic experience curation platform 2730 comprises various components which support and provide capabilities directed to experience curation for platform users 2710. A user management system 2735 is present and configured to provide a portal for each user to create and manage their own user profile. A user profile may be created and stored in a preference database 2736. The user profile may comprise information associated with the user including, but not limited to, user contact information (e.g., name, email address, address, IP address, etc.), demographic information (e.g., age, ethnicity, political party, etc.), login information (e.g., username and password), and a plurality of preferences);
generating, by the processor, a representation of the first information and the natural-language prompt (see [0082], where incoming data 210 represents user data from various sources, such as user interactions, preferences, and external data feeds. This data is crucial for understanding user context, personalizing experiences, and enabling the AI subsystems to make informed decisions. Also see [0108], where the context data 601 is broken into chunks, passed through and embedding model 615, then stored in a specialized database called a vector database 620. Embedding models are a class of models used in many tasks such as natural language processing (NLP) to convert words, phrases, or documents into numerical representations (embeddings) that capture similarity which often correlates semantic meaning. 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. Also see [0109], where a user may submit a query 603 to an experience curation engine 640 which starts the prompt construction and retrieval process. The query is sent to DCG 630 which can send the query to various components such as prompt engineering 625 and embedding model 615. Embedding model 615 receives the query and vectorizes it and stores it in vector database 620. The vector database 620 can send contextual data (via vectors) to DCG 630 and to various APIs/plugins 635. Prompt engineering 625 can receive prompts 602 from developers to train the model on);
querying, by the processor, a second database using the representation to retrieve second information (see [0277], where a vector database 2737 is present and configured to store a plurality of embedded data (i.e., vectors) comprising at least one or more of content requests, un-curated content, extracted and classified content, curated content, device context, and user preferences. A vector database stores vectors, which are mathematical representations of data points in a multi-dimensional space. Each vector typically represents a feature or an entity, and the database is optimized for storing, querying, and manipulating these vectors efficiently. The information stored in vector database 2737 may be leveraged by platform 2730 for use in machine learning, data mining, content extraction and classification, and similarity search processes. Each vector may be associated with a unique identifier and can be of fixed or variable length. Also see [0108], where the vector database 615 is responsible for efficiently storing, comparing, and retrieving a large plurality of embeddings (i.e., vectors));
generating a modified text prompt based on the natural-language prompt, the first information, and the second information (see [0082], where incoming data 210 represents user data from various sources, such as user interactions, preferences, and external data feeds. This data is crucial for understanding user context, personalizing experiences, and enabling the AI subsystems to make informed decisions. Also see [0109], where a user may submit a query 603 to an experience curation engine 640 which starts the prompt construction and retrieval process. The query is sent to DCG 630 which can send the query to various components such as prompt engineering 625 and embedding model 615. Embedding model 615 receives the query and vectorizes it and stores it in vector database 620. The vector database 620 can send contextual data (via vectors) to DCG 630 and to various APIs/plugins 635. Prompt engineering 625 can receive prompts 602 from developers to train the model on. These can include some sample outputs such as in few-shot prompting. The addition of prompts via prompt engineering 625 is designed to ground model responses in some source of truth and provide external context the model wasn't trained on. Other examples of prompt engineering that may be implemented in various embodiments include, but are not limited to, chain-of-thought, self-consistency, generated knowledge, tree of thoughts, directional stimulus, and/or the like. Also see [0272], where the additional content can be fed into a classification model to extract and classify relevant information. User preference data is retrieved. A prompt is engineered using the outputs from the classification tasks performed on both devices and the user preference data);
Crabtree fails to teach transmitting, by the server, the natural-language text output to the user device; and causing, by the user device, the chat application to communicate the natural language text output to the user.
However, Mann does teach:
transmitting, by the server, the natural-language text output to the user device (see [0151], where the selected LLM or other generative engine continues the input prompt and returns that continuation to the caller, which in many cases may be the prompt management service 114. In other cases, output of the generative engine service 116 can be provided to the centralized content service 112 to return to a suitable backend application, to in turn return to or perform a task for the benefit of a client device such as the client device 104 or the client device 106);
and causing, by the user device, the chat application to communicate the natural language text output to the user (see [0175], where the chat service 204 may include a chat-based interface that is incorporated into another graphical user interface or platform frontend or, alternatively, may be a dedicated chat-based platform. Other services that may leverage the system 200 using the intake service 210 include an issue tracking system intake portal, a company directory, a user homepage or other similar interfaces. Also see [0190], where the input is received via the chat service 204, the generative response may be displayed in a reply or message of the chat interface. Also see [0229], where with respect to FIG. 2, a user input 532 provided to the search control 530 (search input control) may result in the display of a generative response 542, which may take the form of a generative answer responsive to an interrogatory of the user input 532. The generative response 542 may be presented in a chat or messaging format in which the user is identified by icon or indicia 524).
Crabtree and Mann are considered to be analogous to the claimed invention because they are in the same field of electric digital data processing. Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Crabtree to incorporate the teachings of Mann by receiving a first user preference to modify a system prompt; providing the modified system prompt as input to a machine-learning language model; receiving a natural-language text prompt from the user; retrieving first information from a first database; generating a representation based on the first information which is used to retrieve second information from a second database; generating a modified text prompt for the purpose of generating a natural-language text output using the natural-language text prompt, the first information, and the second information; and transmitting the output to the user device via a chat application in order to provide a more robust, accurate user experience when interacting with generative artificial intelligence and large language models based on the user’s defined preferences and identity (see Crabtree’s [0009], where what is needed is a new approach to UX design that leverages the power of statistics, machine learning, artificial intelligence, stochastic search, Gen AI, user feedback, and modeling simulation to address the limitations of current UX design, generation, and efficacy modeling systems and methods. Such an approach may enable the creation of agile and adaptive, context-aware, and engaging user experiences that can further the way users interact with digital systems across various domains, content, information, devices, and ongoing sequences of interactions or sessions. Also see Mann’s [0002], where a large amount of user-generated content may be created across multiple platforms. It can be difficult to locate relevant content and even more difficult to synthesize answers to user search queries in an efficient an accurate manner. The systems and techniques described herein may be used to identify and extract relevant content from multiple platforms and present generative and curated results to a user in a generative answer interface).
Regarding claim 2, which depends on claim 1, both Crabtree and Mann teach the limitations in claim 1; furthermore, Crabtree does teach:
Wherein the representation is a vector embedding (see [0108], where the context data 601 is broken into chunks, passed through and embedding model 615, then stored in a specialized database called a vector database 620. Embedding models are a class of models used in many tasks such as natural language processing (NLP) to convert words, phrases, or documents into numerical representations (embeddings) that capture similarity which often correlates semantic meaning. Also see [0146], where 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 second database is a vector database comprising a plurality of vectors (see [0108], where 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. Also see [0146], where 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),
each vector of the plurality of vectors representative of a text segment of a plurality of text segments, and the second information comprises at least one text segment of the plurality of text segments (see [0108], where the context data 601 is broken into chunks, passed through and embedding model 615, then stored in a specialized database called a vector database 620. Embedding models are a class of models used in many tasks such as natural language processing (NLP) to convert words, phrases, or documents into numerical representations (embeddings) that capture similarity which often correlates semantic meaning. Exemplary embedding models can include, but are not limited to, text-embedding-ada-002 model (i.e., OpenAI API), bidirectional encoder representations form transformers, Word2Vec, FastText, transformer-based models, and/or the like. The vector database 615 is responsible for efficiently storing, comparing, and retrieving a large plurality of embeddings (i.e., vectors)).
Regarding claim 5, which depends on claim 1, both Crabtree and Mann teach the limitations in claim 1; furthermore, Crabtree does teach wherein receiving the at least one first operator preference comprises querying, by the server, a first database using the user identifier to retrieve the at least one first operator preference (see [0098], where an enterprise user 410 may refer to a business organization or company. To facilitate the creation of purpose-built, trained model, enterprise user 410 can provide a plurality of enterprise knowledge 411 which can be leveraged to build enterprise specific (or even specific to certain departments within the enterprise) ML/AI models. Enterprise knowledge 411 may refer to documents or other information important for the operation and success of an enterprise. Data from internal systems and databases, such as customer relationship management (CRM) systems, enterprise resource planning (ERP) systems, rules and policies databases, and transactional databases, can provide information about the operational context of an enterprise. For example, product knowledge, market knowledge, industry trends, regulatory knowledge, business processes, customer knowledge, technology knowledge, financial knowledge, organization knowledge, and risk management knowledge may be included in enterprise knowledge base 411. Also see [0271], where a session management system 2733 may be configured to support a user's curated experience between and across multiple user devices 2711. Session management systems 2733 may manage a user session across multiple devices. In some implementations, this may comprise steps involving assigning a unique user identifier (ID) (e.g., username, email address, or generated ID, etc.) to identify the user consistently across devices and sessions; generating a session token (e.g., JSON web token or session ID) when the user logs in or starts a session (e.g., submits a request for content), which may be stored on the client-side (e.g., in cookies or local storage) and sent with each request to identify the session; keeping track of session state and store relevant information such as user ID, session token, device information/state, and session timing data (state and expiration); providing cross-device synchronization which may involve updating session state whenever it changes on one device and propagating those changes to the other devices; and implementing session expiration to ensure sessions are not kept open indefinitely).
Regarding claim 10, which depends on claim 1, both Crabtree and Mann teach the limitations in claim 1; furthermore, Crabtree does teach:
querying, by the processor, a third database with the user identifier to retrieve third information (see [0275], where one or more databases may be present and configured to store a plurality of information obtained from various sources such as users 2710, user devices 2711, and service providers and content providers 2740a-n accessible through a communication network such as the Internet 2740. Some exemplary databases that may be implemented in some embodiments include, but are not limited to, relational databases, key-value stores, document databases, graph databases, vector databases, time-series databases, and column-family stores, to name a few. Also see [0263], where a user management system 2735 is present and configured to provide a portal for each user to create and manage their own user profile. A user profile may be created and stored in a preference database 2736. The user profile may comprise information associated with the user including, but not limited to, user contact information (e.g., name, email address, address, IP address, etc.), demographic information (e.g., age, ethnicity, political party, etc.), login information (e.g., username and password), and a plurality of preferences. A user 2710 may provide this information and their preferences via user management system 2735), and
wherein generating, by the processor, the vector embedding comprises a generating a vector embedding representative of the first information, the third information, and the natural-language prompt (see [0291], where an embedding subsystem 3040 is present and configured to use one or more embedding models to vectorize various types of data obtained and analyzed by platform 2730. A user content request and device context data may be vectorized and stored in a vector database). Also see [0108], where embedding models are a class of models used in many tasks such as natural language processing (NLP) to convert words, phrases, or documents into numerical representations (embeddings) that capture similarity which often correlates semantic meaning).
Regarding claim 11, which depends on claim 1 and claim 10, both Crabtree and Mann teach the limitations in claim 1, and Crabtree teaches the limitations in claim 10; furthermore, Crabtree does teach wherein:
the first database is configured to store data according to a first database management system (see [0263], where a user management system 2735 is present and configured to provide a portal for each user to create and manage their own user profile. A user profile may be created and stored in a preference database 2736. The user profile may comprise information associated with the user including, but not limited to, user contact information (e.g., name, email address, address, IP address, etc.), demographic information (e.g., age, ethnicity, political party, etc.), login information (e.g., username and password), and a plurality of preferences. A user 2710 may provide this information and their preferences via user management system 2735); and
the third database is configured to store data according to a second database management system (see [0275], where one or more databases may be present and configured to store a plurality of information obtained from various sources such as users 2710, user devices 2711, and service providers and content providers 2740a-n accessible through a communication network such as the Internet 2740. Some exemplary databases that may be implemented in some embodiments include, but are not limited to, relational databases, key-value stores, document databases, graph databases, vector databases, time-series databases, and column-family stores, to name a few).
Regarding claim 12, which depends on claim 1, both Crabtree and Mann teach the limitations in claim 1; furthermore, Crabtree does teach the at least one user preference to at least one memory of the user device based on at least one input received by a user interface of the user device, wherein receiving the at least one preference comprises retrieving the at one preference from the at least one memory (see [0278], where preference database 2800 may be stored in local storage on a user device 2711, in a centralized data storage system (e.g., corporate server), in a cloud-based storage provided by platform 2730, or distributed between these or other storage systems. According to the aspect, preference database 2800 may be configured to store a plurality of user preference data 2810-2860. The user preference data may be received, retrieved, or otherwise obtained from a user 2710 and/or their user device(s) 2711. Also see [0263], where a user management system 2735 is present and configured to provide a portal for each user to create and manage their own user profile. A user profile may be created and stored in a preference database 2736. The user profile may comprise information associated with the user including, but not limited to, user contact information (e.g., name, email address, address, IP address, etc.), demographic information (e.g., age, ethnicity, political party, etc.), login information (e.g., username and password), and a plurality of preferences. A user 2710 may provide this information and their preferences via user management system 2735. A user 2710 may provide this information to platform 2730 via a website or web application).
Regarding claim 16, which depends on claim 1, both Crabtree and Mann teach the limitations in claim 1; furthermore, Crabtree does teach requesting, by the server, the at least one user preference from the user device upon receiving the natural-language text prompt (see [0123] where a response portal 1020 is present and configured to receive a response from one a Gen AI model and a response management system 1030 determines if the received response needs to be curated or not. If the response does not need to be curated, then it may be sent as an uncrated response 1002 to the user who submitted the query. Response management 1030 can determine if there are any user/entity defined rules or preferences available such as stored in a user/entity profile in a data storage system of platform 420. Rules management 1040 can retrieve said rules and response management can curate or otherwise augment the received response based on the user/entity rules or preferences. The result is a curated response 1002 which can be transmitted back to the user who submitted the query. Also see [0263], where a user profile may be created and stored in a preference database 2736. The user profile may comprise information associated with the user including, but not limited to, user contact information (e.g., name, email address, address, IP address, etc.), demographic information (e.g., age, ethnicity, political party, etc.), login information (e.g., username and password), and a plurality of preferences. A user 2710 may provide this information and their preferences via user management system 2735. Also see [0264], where a context extraction and tracking system 2731 is present and configured to receive the content request from a user device 2711, format the request and submit it to the service provider/content provider on the user's behalf, receive un-curated content from the service provider/content provider, and extract and classify relevant content from the received un-curated content).
Regarding claim 17, which depends on claim 1 and claim 16, both Crabtree and Mann teach the limitations in claim 1, and Crabtree teaches the limitations in claim 16; furthermore, Crabtree does teach:
wherein querying the first database comprises querying the first database upon receiving the natural-language text prompt (see [0276], where a preference database 2736 is present and configured to store a user profile and a plurality of user preferences which may be explicitly defined/declared by the user. Also see [0232] where the responses to wizard/chatbot, the design elements and/or templates selected by the user, the user’s interaction with the workboard (e.g., search queries, mouse clicks, hover time, etc.), and any available user preferences (e.g., retrieved from a preference database or submitted directly by the user) may be included in a user specification. Also see [0123], where response management 1030 can determine if there are any user/entity defined rules or preferences available such as stored in a user/entity profile in a data storage system of platform 420. Rules management 1040 can retrieve said rules and response management can curate or otherwise augment the received response based on the user/entity rules or preferences. The result is a curated response 1002 which can be transmitted back to the user who submitted the query) and
wherein querying the second database comprises querying the second database upon receiving the natural-language text prompt (see [0277], where a vector database 2737 is present and configured to store a plurality of embedded data (i.e., vectors) comprising at least one or more of content requests, un-curated content, extracted and classified content, curated content, device context, and user preferences. A vector database stores vectors, which are mathematical representations of data points in a multi-dimensional space. Each vector typically represents a feature or an entity, and the database is optimized for storing, querying, and manipulating these vectors efficiently. The information stored in vector database 2737 may be leveraged by platform 2730 for use in machine learning, data mining, content extraction and classification, and similarity search processes. Each vector may be associated with a unique identifier and can be of fixed or variable length).
Regarding claim 18, which depends on claim 1, claim 16, and claim 17, both Crabtree and Mann teach the limitations in claim 1, and Crabtree teaches the limitations in both claim 16 and claim 17; furthermore, Crabtree does teach wherein generating the at least one user preference comprises generating a first natural-language word based on the at least one input (see [0245], where a design clarification subsystem 2206 may compute a specification clarity score 2610 based on multiple factors including but not limited to, one or more defined goals 2601, specific context 2602, a level of specificity 2603, available examples 2604, and the user's natural language 2605. A general approach to crafting a prompt for Gen AI system may involve obtaining a clearly defined goal of the prompt, or in other words, what the user wants the Gen AI system to generate. This could be a text description, a piece of code, an image, or any other type of content. Specific context may further comprise user preferences).
Regarding claim 19, which depends on claim 1, claim 16, and claim 17, both Crabtree and Mann teach the limitations in claim 1, and Crabtree teaches the limitations in both claim 16 and claim 17; furthermore, Crabtree does teach wherein the at least one first operator preference comprises a second natural-language word (see [0098], where enterprise knowledge 411 may refer to documents or other information important for the operation and success of an enterprise. Data from internal systems and databases, such as customer relationship management (CRM) systems, enterprise resource planning (ERP) systems, rules and policies databases, and transactional databases, can provide information about the operational context of an enterprise. For example, product knowledge, market knowledge, industry trends, regulatory knowledge, business processes, customer knowledge, technology knowledge, financial knowledge, organization knowledge, and risk management knowledge may be included in enterprise knowledge base 411).
Claim(s) 3, and 6 is/are rejected under 35 U.S.C. 103 as being unpatentable over Crabtree (US 20250258685 A1) in view of Mann (US 20250217576 A1) and further in view of Li (US 20250141769 A1).
Regarding claim 3, which depends on claim 1 and claim 2, both Crabtree and Mann teach the limitations in claim 1, and Crabtree teaches the limitations of claim 2; however, Crabtree and Mann fail to teach receiving at least one first operator preference indicative of at least one second characteristic, preferred by an operator of the server, of the natural-language outputs, wherein modifying the system prompt comprises modifying the system prompt based on the received at least one user preference and the at least one first operator preference to generate the modified system prompt.
However, Li does teach:
receiving at least one first operator preference indicative of at least one second characteristic, preferred by an operator of the server, of the natural-language outputs (see [0048], where the customer premises equipment (CPE) 115 includes a local chat prompt and response engineering (CPRE) component 205a, which can comprise, e.g., software running on a processor of the CPE 115, as well as a local adaptive multi-modal ML engine (ML Engine 210a), which can also comprise, e.g., firmware or software running on a processor of the CPE 115. The local CPRE 205a can include a prompt and response cache 225. The CPRE 205a also has access to (and/or can collect, receive and/or store) input data 230, which can include without limitation, platform profile and status information 230 of the CPE 115 and/or other local conditions (e.g., local network statistics and information, weather data, etc.), as well as user and/or operator preferences 235).
wherein modifying the system prompt comprises modifying the system prompt based on the received at least one user preference and the at least one first operator preference to generate the modified system prompt (see [0035], where some embodiments allow the seamless integration of a chat engine (or chat bot) into the analytic system to generate comprehensive and insightful information for the operator and the user, with chat prompts and responses being engineered (e.g., enhanced, augmented, etc. based on the techniques described below) to adapt to the dynamic conditions of the user input, device and network status, user/operator preferences, and the like).
Crabtree, Mann, and Li are considered to be analogous to the claimed invention because they are in the same field of electric digital data processing. Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Crabtree and Mann to incorporate the teachings of Li to receive at least one preferred first operator preference indicative of at least one second characteristic of the natural-language outputs, and modifying and generating a system prompt based on the received at least one user preference in order to improve the response accuracy and adaptability based on operator knowledge and preferences (see [0022], where various embodiments can utilize diagnostics and/or operations information of device platforms and networks to improve chat accuracy and relevancy via chat prompt enrichment and chat response augmentation. Also see [0035], where various embodiments can provide substantial improvements to current systems and processes. Merely by way of example, some embodiments allow the seamless integration of a chat engine (or chat bot) into the analytic system to generate comprehensive and insightful information for the operator and the user, with chat prompts and responses being engineered (e.g., enhanced, augmented, etc. based on the techniques described below) to adapt to the dynamic conditions of the user input, device and network status, user/operator preferences, and the like).
Regarding claim 6, which depends on claim 1, both Crabtree and Mann teach the limitations in claim 1; however, Crabtree and Mann fail to teach receiving at least one second operator preference indicative of at least one third characteristic, preferred by the operator of the server, of the natural-language outputs, wherein modifying the system prompt comprises modifying the system prompt based on the on the retrieved at least one user preference, the at least one first operator preference, and the at least one second operator preference to generate the modified system prompt.
However, Li does teach:
receiving at least one second operator preference indicative of at least one third characteristic, preferred by the operator of the server, of the natural-language outputs (see [0048], where the local chat prompt and response engineering (CPRE) 205a also has access to (and/or can collect, receive and/or store) input data 230, which can include without limitation, platform profile and status information 230 of the CPE 115 and/or other local conditions (e.g., local network statistics and information, weather data, etc.), as well as user and/or operator preferences 235),
wherein modifying the system prompt comprises modifying the system prompt based on the on the retrieved at least one user preference, the at least one first operator preference, and the at least one second operator preference to generate the modified system prompt (see [0052], where the local chat prompt and response engineering (CPRE) function 205a can perform local prompt composition and enrichment using the ML-processed user prompt, based, in some cases, on the data 230 and/or the user and operator preferences 235. ⋅ The local CPRE function 205a can also perform local response augmentation using the response received one or more of the remote devices 201, in some cases by augmenting the response based on the data 230 and/or the user and operator preferences 235. Also see [0057], where the local CPRE function 205a generates an enhanced prompt, e.g., by forming an original prompt from the raw user query and/or the input from the ML Engine 210a and/or enriches the prompt to provide an enriched prompt, e.g., by incorporating prompt parameters (which can include, without limitation, the user and operator preferences 235 and the device platform profile and status 230) with the user query/original prompt).
Crabtree, Mann, and Li are considered to be analogous to the claimed invention because they are in the same field of electric digital data processing. Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Crabtree and Mann to incorporate the teachings of Li to receive at least one preferred second operator preference indicative of at least one third characteristic of the natural-language outputs, and modifying and generating a system prompt based on the on the retrieved at least one user preference, the at least one first operator preference, and the at least one second operator in order to improve the accuracy of the response based on user and operator knowledge, preferences, and context (see [0044], where the chat engine 135 can receive a prompt (e.g., a prompt formed from a user query, as described herein) and provide a response. As described herein, various embodiments can enhance the prompt and/or the response to increase the likelihood that the response will include actionable, understandable information that properly addresses the (real or perceived) condition that provoked the user query).
Claim(s) 20 is/are rejected under 35 U.S.C. 103 as being unpatentable over Crabtree (US 20250258685 A1) in view of Li (US 20250141769 A1) and further in view of Mann (US 20250217576 A1).
Regarding claim 20, Crabtree teaches the following limitations:
A system for natural language generation, the system comprising: a first database configured to store first user-specific information (see [0276], where a preference database 2736 is present and configured to store a user profile and a plurality of user preferences which may be explicitly defined/declared by the user and/or implicitly learned/derived from user behavior and interactions);
a second database configured to store a plurality of vector embeddings representative of a plurality of natural-language text segments, each vector embedding of the plurality of vector embeddings representative of one natural-language text segment of the plurality of natural-language text segments (see [0277], where a vector database 2737 is present and configured to store a plurality of embedded data (i.e., vectors) comprising at least one or more of content requests, un-curated content, extracted and classified content, curated content, device context, and user preferences. A vector database stores vectors, which are mathematical representations of data points in a multi-dimensional space. Each vector typically represents a feature or an entity, and the database is optimized for storing, querying, and manipulating these vectors efficiently. The information stored in vector database 2737 may be leveraged by platform 2730 for use in machine learning, data mining, content extraction and classification, and similarity search processes. Each vector may be associated with a unique identifier and can be of fixed or variable length);
a user device comprising: a first processor (see [0301], where a user device 3210a may comprise an experience curation application 3211a which may be a web application or a mobile device software application, one or more filters (e.g., Gen AI models), and a local instance of a preference database 3213a. User device 3210a may further comprise an operating system, one or more processors, a memory, a display, input device(s) (e.g., mouse, keyboard, touchscreen, etc.), embedded sensors (e.g., microphone, camera, fingerprint sensor, facial recognition system, gyroscope, Lidar, gait detection, dental record comparisons, implant/fixation device comparisons, tattoo comparisons, etc.), and other applications or microservices); and
at least one first memory encoded with first instructions that, when executed, cause the first processor to: receive at least one input indicative of a natural-language text string; and provide the natural-language text string as a natural-language text prompt to a chat application operating on the user device (see [0081] The chatbot AI 203 powers the conversational interfaces within the system, enabling natural language interactions between the user and the AI. It may utilize natural language processing (NLP) and natural language generation (NLG) techniques to understand user inputs, maintain context, and provide intelligent and personalized responses. It may further utilize retrieval augmented generation (RAG), hyperparameter optimization, knowledge graphs and vector representations and databases. The UI/UX AI 204 focuses on the design and optimization of the user interface and user experience elements. It takes into account user preferences, behavior patterns, and contextual information to generate intuitive, user-friendly, and visually intentional interfaces that adapt to the user's needs, process needs or creators' desires); and
a remote device communicatively connected to the user device, the remote device comprising: a second processor (see [0093], where FIG. 4 is a block diagram illustrating an exemplary system architecture for a distributed generative artificial intelligence reasoning and action platform 420, according to an embodiment. According to the embodiment, platform 420 is configured as a cloud-based computing platform comprising various system or sub-system components configured to provide functionality directed to the execution of neuro-symbolic Gen AI reasoning and action. Exemplary platform systems can include a distributed computational graph (DCG) computing system 421, a curation computing system 422, a marketplace computing system 423, and a context computing system 424. In some embodiments, systems 421-424 may each be implemented as standalone software applications or as a services/microservices architecture which can be deployed (via platform 420) to perform a specific task or functionality. In such an arrangement, services can communicate with each other over an appropriate network using lightweight protocols such as HTTP, gRPC, or message queues. This allows for asynchronous and decoupled communication between services. Services may be scaled independently based on demand, which allows for better resource utilization and improved performance. Services may be deployed using containerization technologies such as Docker or containerd and orchestrated using container orchestration platforms like Kubernetes. This allows for easier deployment and management of services); and
at least one second memory encoded with second instructions that, when executed, cause the second processor to: receive the natural language text prompt from the user device (see [0264], where a context extraction and tracking system 2731 is present and configured to receive the content request from a user device 2711, format the request and submit it to the service provider/content provider on the user's behalf, receive un-curated content from the service provider/content provider, and extract and classify relevant content from the received un-curated content. Also see [0122], where a query portal 1010 may be present and configured to receive a query 1001 and prepare it for processing by a Gen AI model);
receive at least one user preference indicative of at least one first characteristic, preferred by a user, of natural-language outputs generated by a machine-learning language model based on user provided natural-language text inputs (see [0083], where a user preference manager 230 enables users to set and modify their preferences, including accessibility settings, content preferences, and personalization options. Also see [0011], where a system for user experience curation, comprising: a computing device comprising at least a memory and a processor; a plurality of programming instructions that, when operating on the processor, cause the computing device to: collect a plurality of data types from a plurality of sources or systems wherein data may include but is not limited to audio, visual, telematics, haptic, smells, environmental conditions, user security settings, user privacy settings, user disabilities and sensory ranges, user health and faculties, other user devices, and user preference settings);
provide the system prompt as an initial input to the machine-learning language model (see [0109], where prompt engineering 625 can receive prompts 602 from developers to train the model on. Also see [0110], where during a prompt execution process, experience curation 640 can send user query to DCG 630 which can orchestrate the retrieval of context and a response. Using its declarative roots, DCG 630 can abstract away many of the details of prompt chaining; interfacing with external APIs 635 (including determining when an API call is needed); retrieving contextual data from vector databases 630; and maintaining memory across multiple LLM calls. The DCG output may be a prompt, or series of prompts, to submit to a language model via LLM services);
query the first database with the user identifier to retrieve first information (see [0263], where a user management system 2735 is present and configured to provide a portal for each user to create and manage their own user profile. A user profile may be created and stored in a preference database 2736. The user profile may comprise information associated with the user including, but not limited to, user contact information (e.g., name, email address, address, IP address, etc.), demographic information (e.g., age, ethnicity, political party, etc.), login information (e.g., username and password), and a plurality of preferences. A user 2710 may provide this information and their preferences via user management system 2735. A user 2710 may provide this information to platform 2730 via a website or web application accessible via an appropriate Internet browser operating on a user device 2711 or by using a mobile device experience curation software application (App) to access platform 2730. Also see [0271], where in some implementations, this may comprise steps involving assigning a unique user identifier (ID) (e.g., username, email address, or generated ID, etc.) to identify the user consistently across devices and sessions);
generate a vector embedding representative of the first information and the natural-language prompt (see [0291], where an embedding subsystem 3040 is present and configured to use one or more embedding models to vectorize various types of data obtained and analyzed by platform 2730. A user content request and device context data may be vectorized and stored in a vector database. Also see [0108], where the context data 601 is broken into chunks, passed through and embedding model 615, then stored in a specialized database called a vector database 620. Also see [0272], where the additional content can be fed into a classification model to extract and classify relevant information. User preference data is retrieved. A prompt is engineered using the outputs from the classification tasks performed on both devices and the user preference data. The prompt and content elements are submitted to Gen AI models. The output of the Gen AI models is rendered on the user interface of the second device as curated content responsive to the content request);
query the second database using the vector embedding to retrieve second information (see [0277], where a vector database 2737 is present and configured to store a plurality of embedded data (i.e., vectors) comprising at least one or more of content requests, un-curated content, extracted and classified content, curated content, device context, and user preferences. A vector database stores vectors, which are mathematical representations of data points in a multi-dimensional space. Each vector typically represents a feature or an entity, and the database is optimized for storing, querying, and manipulating these vectors efficiently. The information stored in vector database 2737 may be leveraged by platform 2730 for use in machine learning, data mining, content extraction and classification, and similarity search processes. Each vector may be associated with a unique identifier and can be of fixed or variable length. Also see [0108], the vector database 615 is responsible for efficiently storing, comparing, and retrieving a large plurality of embeddings (i.e., vectors));
generate a modified text prompt based on the natural-language prompt, the first information, and the second information (see [0265], where a process filtering system 2732 is present and configured to obtain a plurality of information comprising at least user preference data, classified content, and context data and to use the obtained plurality of information as inputs to create one or more filters which can be used to curate the user's experience when browsing the web. Also see [0236], where based on the task, prompt engineering subsystem 2402 designs a prompt that provides the necessary context for the agent(s));
provide, subsequent to providing the system prompt, the modified text prompt as an input to the machine-learning language model to generate a natural-language text output (see [0110], the DCG output may be a prompt, or series of prompts, to submit to a language model via LLM services 660. The DCG output may be a prompt, or series of prompts, to submit to a language model via LLM services 660 (which may be potentially prompt tuned). 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).
Crabtree fails to teach at least one second memory encoded with second instructions that, when executed, cause the second processor to: receive at least one first operator preference indicative of at least one second characteristic, preferred by an operator of the server, of the natural language outputs; modify a system prompt for the machine-learning language model based on the at least one user preference and the at least one first operator preference; and transmit the natural-language text output to the user device.
However, Li does teach at least one second memory encoded with second instructions that, when executed, cause the second processor to:
receive at least one first operator preference indicative of at least one second characteristic, preferred by an operator of the server, of the natural language outputs (see [0048], where the local customer premises equipment (CPE) 115 includes a local chat prompt and response engineering (CPRE) component 205a, which can comprise, e.g., software running on a processor of the CPE 115, as well as a local adaptive multi-modal ML engine (ML Engine 210a), which can also comprise, e.g., firmware or software running on a processor of the CPE 115. The local CPRE 205a can include a prompt and response cache 225. The CPRE 205a also has access to (and/or can collect, receive and/or store) input data 230, which can include without limitation, platform profile and status information 230 of the CPE 115 and/or other local conditions (e.g., local network statistics and information, weather data, etc.), as well as user and/or operator preferences 235);
modify a system prompt for the machine-learning language model based on the at least one user preference and the at least one first operator preference; and transmit the natural-language text output to the user device (see [0035], where merely by way of example, some embodiments allow the seamless integration of a chat engine (or chat bot) into the analytic system to generate comprehensive and insightful information for the operator and the user, with chat prompts and responses being engineered (e.g., enhanced, augmented, etc. based on the techniques described below) to adapt to the dynamic conditions of the user input, device and network status, user/operator preferences, and the like).
Crabtree and Li both fail to teach at least one second memory encoded with second instructions that, when executed, cause the second processor to: transmit the natural-language text output to the user device.
However, Mann does teach at least one second memory encoded with second instructions that, when executed, cause the second processor to: transmit the natural-language text output to the user device (see [0151], where the selected LLM or other generative engine continues the input prompt and returns that continuation to the caller, which in many cases may be the prompt management service 114. In other cases, output of the generative engine service 116 can be provided to the centralized content service 112 to return to a suitable backend application, to in turn return to or perform a task for the benefit of a client device such as the client device 104 or the client device 106).
Crabtree, Mann, and Li are considered to be analogous to the claimed invention because they are in the same field of electric digital data processing. Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Crabtree to incorporate the teachings of Mann and Li to include a system with first and second processors capable of receiving a first user preference and a first operator preference to modify a system prompt; providing the modified system prompt as input to a machine-learning language model; receiving a natural-language text prompt from the user; retrieving first information from a first database; generating a vector representation based on the first information which is used to retrieve second information from a second database; generating a modified text prompt for the purpose of generating a natural-language text output using the natural-language text prompt, the first information, and the second information; and transmitting the output to the user device via a chat application in order to provide a system with a more robust, accurate user experience when interacting with generative artificial intelligence and large language models based on the user and operator’s defined preferences (see Crabtree’s [0009], where what is needed is a new approach to UX design that leverages the power of statistics, machine learning, artificial intelligence, stochastic search, Gen AI, user feedback, and modeling simulation to address the limitations of current UX design, generation, and efficacy modeling systems and methods. Such an approach may enable the creation of agile and adaptive, context-aware, and engaging user experiences that can further the way users interact with digital systems across various domains, content, information, devices, and ongoing sequences of interactions or sessions. Also see Mann’s [0002], where a large amount of user-generated content may be created across multiple platforms. It can be difficult to locate relevant content and even more difficult to synthesize answers to user search queries in an efficient an accurate manner. The systems and techniques described herein may be used to identify and extract relevant content from multiple platforms and present generative and curated results to a user in a generative answer interface).
Claim(s) 4 is/are rejected under 35 U.S.C. 103 as being unpatentable over Crabtree (US 20250258685 A1) in view of Mann (US 20250217576 A1) and further in view of Goel (US 20240168978 A1).
Regarding claim 4, which depends on claim 1, both Crabtree and Mann teach the limitations in claim 1; however, both Crabtree and Mann fails to teach wherein: the vector database comprises a plurality of partitions of vector data; querying the second database using the representation comprises: selecting a partition of vector data of the plurality of partitions of vector data based on the user identifier; and comparing the vector embedding to vectors of the partition of vector data to retrieve the second information.
However, Goel does teach wherein:
the vector database comprises a plurality of partitions of vector data (see [0023], where the vector database provides fast and efficient billion scale approximate nearest neighbor search for multiple tenants. The indexes are partitioned on the basis of tenant ID to enable multi-tenancy and can be scaled to millions of organizations with each organization containing millions of vectors);
querying the second database using the representation comprises: selecting a partition of vector data of the plurality of partitions of vector data based on the user identifier (see [0020], where the present subject matter can be implemented for providing nearest neighbor vectors from in the vector database to a tenant in a multi-tenant environment, for instance, when the vector database has been indexed in the manner described previously. A user may make a request to a service to provide the nearest neighbor vectors from the index of the tenant. The request may include a query vector and input parameters indicating the number of objects to be read to fulfill the request of the user. The request may be, for example, “For a given vector, can you find me the nearest neighbor vectors for a given tenant and a given index?”); and
comparing the vector embedding to vectors of the partition of vector data to retrieve the second information (see [0014] where generally, in conventional scenarios, for a nearest neighbor query, the vector database is preprocessed to create an index for querying. For a given query vector, the created index is used to identify a set of vectors that are likely to be close to the query vector. In an example, when a vector database receives a query, the vector database compares the indexed vectors to the query vector to determine the nearest vector neighbors).
Crabtree, Mann, and Goel are considered to be analogous to the claimed invention because they are in the same field of electric digital data processing. Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Crabtree and Mann to incorporate the teachings of Crabtree to include partitions of vector data in the vector database; and querying and comparing the vector embeddings to vectors of the partitions to retrieve the second information in order to enable efficient, scalable, and isolated multi-tenant search at a low computational and memory cost to retrieve the second information (see [0023], where an advantage of the present system is that it does not incur any cost (beyond a minimal storage cost) for an index when the corresponding tenant is not utilizing the index. At the same time, the index can provide sub-100 ms response when an index is used for the first time or after a long period of inactivity. Also see [0024], where unlike conventional vector database solutions which does not scale in a multi-tenant environment and requires significant amount of memory associated with high costs, the present subject matter is a 100% on-disk solution, thereby ensuring that the indexes can be scaled to millions of tenants while keeping the memory costs minimal).
Claim(s) 13 is/are rejected under 35 U.S.C. 103 as being unpatentable over Crabtree (US 20250258685 A1) in view of Mann (US 20250217576 A1) and further in view of Shamir (US 20250292098 A1).
Regarding claim 13, which depends on claim 1 and claim 12, both Crabtree and Mann teach the limitations in claim 1, and Crabtree teaches the limitations in claim 12; however, both Crabtree and Mann fails to teach wherein: encoding the at least one user preference to the at least one memory comprises: generating at least one token representative of the at least one user preference using a tokenizer algorithm configured to generate input tokens usable by the machine-learning language model; and storing the at least one token to the at least one memory, and retrieving the at least one user preference comprises retrieving the at least one token.
However, Goel does teach wherein:
encoding the at least one user preference to the at least one memory comprises: generating at least one token representative of the at least one user preference using a tokenizer algorithm configured to generate input tokens usable by the machine-learning language model (see [0032], where the present disclosure provides a framework for fine-tuning pre-trained sequence processing models to human preferences and/or other objective(s). The model can be tuned to predict posterior token probabilities conditioned on the human preferences. Also see [0044], where LLMs can be further fine-tuned to specific preferences, that are either human preferences or preferences aligned with specialized target or control tasks. Reinforcement learning with human feedback (RLHF) has emerged as a method for fine-tuning LLMs to human preferences. A set of prompts is provided to the pre-trained model. For each prompt, multiple token response sequences are sequentially sampled from the distributions predicted by the base model. Human labelers rate or rank the multiple sequences, and a reward model is trained with embedding inputs representing the sampled token sequences to model the human preference among the multiple sampled sequences. The trained reward model is then iteratively used in a reinforcement learning loop to fine-tune the model towards the human preferences. Also see [0200], where for example, elements 5-1, 5-2, . . ., 5-M can represent tokens obtained using a tokenizer. For instance, a tokenizer can process a given portion of an input source and output a series of tokens that represent the portion of the input source); and
storing the at least one token to the at least one memory, and retrieving the at least one user preference comprises retrieving the at least one token (see [0249], where model host 31 can access runtime data source(s) 37 for augmenting input(s) 2 with additional contextual information. For instance, runtime data source(s) 37 can include a knowledge graph 37-1 that facilitates structured information retrieval for information associated with input request(s) 33 (e.g., a search engine service). Runtime data source(s) 37 can include public or private, external or local database(s) 37-2 that can store information associated with input request(s) 33 for augmenting input(s) 2. Runtime data source(s) 37 can include account data 37-3 which can be retrieved in association with a user account corresponding to a client 32 for customizing the behavior of model host 31 accordingly).
Crabtree, Mann, and Shamir are considered to be analogous to the claimed invention because they are in the same field of electric digital data processing. Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Crabtree and Mann to incorporate the teachings of Shamir to generate at least one token representative of the at least one user preference using a tokenizer algorithm and store them to memory for future use when retrieving a user preference in order to better personalize and/or customize the user’s data preference for better flexibility (see [0033], where the proposed methods are universal in that they allow: fine-tuning to any type of reward (or preference) objective; tuning to simultaneously satisfy multiple “control” preferences addressing limitations of other methods; and/or the same model to be tuned to simultaneously support multiple different “expert” tasks. The proposed techniques can be applied to fine-tune and deploy a single model, while preserving the ability to generate predictions of the base model if desired).
Allowable Subject Matter
Claim(s) 7-9, and 14-15 objected to as being dependent upon a rejected base claim, but would be allowable if rewritten in independent form including all of the limitations of the base claim and any intervening claims.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to JOHN HONG FANG-WU whose telephone number is (571)270-0607. The examiner can normally be reached Monday - Friday, 8AM to 5PM.
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 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.
/JOHN HONG FANG-WU/ Examiner, Art Unit 2653
/Paras D Shah/ Supervisory Patent Examiner, Art Unit 2653
07/06/2026
Prosecution Insights