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 .
Priority
Acknowledgment is made of applicant’s claim for foreign priority under 35 U.S.C. 119 (a)-(d). The certified copy has been filed in parent Application No. CN202411873277.5, filed on 12/18/2024.
Response to Amendment
This action is in response to the Pre-Brief Appeal Conference decision mailed 6/16/2026 wherein the conferences determined that prosecution would be reopened. Therefore: THIS ACTION IS MADE NON-FINAL.
Status of Claims
Claims 1-20 are currently pending in the present application.
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 and 11 is/are rejected under 35 U.S.C. 103 as being unpatentable over Doggett et al. (US PGPUB No. 2024/0386217; Pub. Date: Nov. 21, 2024) in view of Gelfenbeyn et al. (US PGPUB No. 2024/0221269; Pub. Date: Jul. 4, 2024).
Regarding independent claim 1,
Doggett discloses a generative question answering system for generating style texts, comprising: an input-output device, configured to receive input information; See FIG. 1 & Paragraph [0016], (Disclosing a system for assessing a suitability of a response of an AI character. FIG. 1 illustrates a computing platform 102 having input unit 130 which allows users to interact with AI characters 116a-116b, i.e. a generative question answering system for generating style texts, comprising: an input-output device, configured to receive input information (e.g. input unit 130 receives input from a human user);)
a memory, configured to store a character database and a text knowledge database, See Paragraph [0016], (Disclosing a system for assessing a suitability of a response of an AI character. Character profile database 120 may include character profiles 122a-122c and one or more trained ML models 128 including one or more large language models.) See Paragraph [0047], (The system may determine in-world speech for an AI character based on an original source vocabulary comprising language included in the creative or historical corpus portraying a particular character, i.e. a memory, configured to store a character database (e.g. character profile database 120) and a text knowledge database (e.g. the original source vocabulary associated with each character comprises text).)
wherein the character database stores multiple character templates comprising multiple character descriptions and multiple dialogue examples corresponding to the multiple character templates, See Paragraph [0016], (Character profile database 120 includes character profiles 122a-122c and one or more trained ML models 128 including one or more large language models.) See FIG. 3A & Paragraph [0042], (FIG. 3A illustrates a display pane 300A describing a character profile of Alice from the film Alice in Wonderland including a character description and dialogue history. Note [0020] wherein character profile database 120 may typically store a plurality of character profiles, i.e. wherein the character database stores multiple character templates comprising multiple character descriptions (e.g. character profile database 120 includes a plurality of profiles) and multiple dialogue examples corresponding to the multiple character templates (e.g. profiles are associated with a dialogue history).)
wherein the text knowledge database stores multiple candidate texts; See Paragraph [0047], (The system may determine in-world speech for an AI character based on an original source vocabulary comprising language included in the creative or historical corpus portraying a particular character, i.e. a memory, configured to store a character database (e.g. character profile database 120) and a text knowledge database (e.g. the original source vocabulary associated with each character comprises text).)
wherein the first output text is a natural language response corresponding to the input information; See FIG. 4 & Paragraph [0052], (Flowchart 470 illustrates a method comprising step 471 of receiving dialogue data identifying a character, a storyline and a speech for the character intended to advance the storyline or achieve a goal of the speech. Note [0015] wherein AI characters may be represented by text, audio or both and exhibit characteristics of living, historical or fictional characters in a manner that are recognizable by humans as a personality.) See FIG. 3A, (FIG. 3A illustrates a graphical user interface including a dialogue history of an AI character including text responses that are relevant to a user's initial inquiry, i.e. wherein the first output text is a natural language response corresponding to the input information;)
obtaining, from the character database, a first character template and at least one of the multiple dialogue examples corresponding to the first character template based on the input information; See FIG. 5, (FIG. 5 illustrates method 580 wherein the system may utilize an ML model, such as an LLM (see [0064]) to infer an intended personality profile to be used to advance a storyline or achieve a goal via speech. See Paragraph [0047], (A character may be associated with an original source vocabulary used in to portray a particular character, i.e. obtaining, from the character database, a first character template (e.g. character profiles are stored in character profile database 120) and at least one of the multiple dialogue examples corresponding to the first character template based on the input information (e.g. the original source vocabulary is used to generate in-character interactions);)
Doggett does not disclose a processor, connected to the memory and the input-output device, and configured to perform processes of: obtaining at least one of the multiple candidate texts from the text knowledge database based on the input information, and generating a first output text by processing the input information and the at least one of the multiple candidate texts using a large language model, wherein the first output text is a natural language response corresponding to the input information;
and generating a second output text by transferring the first output text into a specific speaking style defined by the first character template based on the input information, the first character template, and the at least one of the multiple dialogue examples, and the first output text.
Gelfenbeyn discloses a processor, connected to the memory and the input-output device, and configured to perform processes of: obtaining at least one of the multiple candidate texts from the text knowledge database based on the input information, and generating a first output text by processing the input information and the at least one of the multiple candidate texts using a large language model, wherein the first output text is a natural language response corresponding to the input information; See Paragraph [0032], (Disclosing a system for providing contextually oriented behavior of AI characters. The method receives existing knowledge provided by a user and fetches knowledge from a database as well as game engine information and scene information to create a contextual input for an AI character based on the received information.) See Paragraph [0040], (An AI character model may retrieve information from one or more data sources in order to interact with a user in the virtual environment, i.e. obtaining at least one of the multiple candidate texts from the text knowledge database based on the input information (e.g. an AI character may receive inputs and retrieve information from one or more data sources).) See FIG. 5 & Paragraph [0066], (FIG. 5 illustrates an architecture of an AI character model comprising Step B which includes Step 506 of pre-processing incoming data streams comprising multimodal inputs at a client. At step 508, the method generates intermediate outputs using machine learning models representing different elements of cognition, i.e. generating a first output text by processing the input information and the at least one of the multiple candidate texts using a large language model (e.g. the AI character model generates an intermediate output), wherein the first output text is a natural language response corresponding to the input information (e.g. the intermediate outputs represent different elements of cognition and are generated based on pre-processed incoming data stream data);)
and generating a second output text by transferring the first output text into a specific speaking style defined by the first character template based on the input information, the first character template, and the at least one of the multiple dialogue examples, and the first output text. See FIG. 5 & Paragraphs [0074]-[0075], (At Step C of method 500, the system may perform step 510 of composing intermediate outputs into templated formats for ingestion by large language models, animation, gesture and action systems. Step D then uses primary models and systems to generate final behavior-aligned data outputs, i.e. generating a second output text by transferring the first output text into a specific speaking style defined by the first character template (e.g. by generating behavior-aligned data outputs) based on the input information (e.g. Note [0064] describing Step A of FIG. 5 comprising 504 streaming multimodal inputs) , the first character template (e.g. Note [0074] describing Orchestration Step C of FIG. 5 comprising feeding intermediate outputs into templates for ingestion by LLMs and animation, gesture and action models), and the at least one of the multiple dialogue examples, and the first output text (e.g. intermediate outputs generated at step 508 are fed into the templates).)
The examiner notes that Gelfenbeyn discloses a multi-step method of receiving an input, generating an intermediate output, providing said intermediate output to a template and generating in-character outputs. While Gelfenbeyn does not disclose the use of "multiple dialogue examples" to generate an output, Doggett is relied upon to disclose the "multiple dialogue examples" of the claimed invention via the "original source vocabulary" concept described in at least Paragraphs [0046]-[0047] & [0070]. Therefore, one of ordinary skill in the art would recognize that the combination of Gelfenbeyn and Doggett would result in a template that may incorporate additional types of data when generating a response, such as by including elements of an "original source vocabulary" to ensure that a generated output speech is consistent with character parameters.
Doggett and Gelfenbeyn are analogous art because they are in the same field of endeavor, conversational LLMs. It would have been obvious to anyone having ordinary skill in the art before the effective filing date to modify the system of Doggett to include the multi-step process for generating and providing AI characters as illustrated in FIG. 5 of Gelfenbeyn. Paragraph [0122] of Gelfenbeyn discloses that the method allows AI characters to share information across each other to provide contextually oriented behavior, thereby improving the user experience in the virtual environment.
Regarding independent claim 11,
The claim is analogous to the subject matter of independent claim 1 directed to a method or process and is rejected under similar rationale.
Claim(s) 2 and 12 is/are rejected under 35 U.S.C. 103 as being unpatentable over Doggett in view of Gelfenbeyn as applied to claim 1 above, and further in view of Cui et al. (US PGPUB No. 2024/0411790; Pub. Date: Dec. 12, 2024).
Regarding dependent claim 2,
As discussed above with claim 1, Doggett-Gelfenbeyn discloses all of the limitations.
Doggett-Gelfenbeyn does not disclose the step wherein the processor is further configured to perform processes of: computing multiple similarities between a user input content of the input information and the multiple candidate texts of the text knowledge database;
and obtaining the at least one of the multiple candidate texts based on the multiple similarities.
Cui discloses the step wherein the processor is further configured to perform processes of: computing multiple similarities between a user input content of the input information and the multiple candidate texts of the text knowledge database; See Paragraphs [0066]-[0067], (Disclosing a system for answer information generation based on a large language model. The system may obtain candidate documents from a document library based on at least two of the semantic vector of the question text, the at least one piece of argument information, and the event category. Documents having the highest semantic similarities are obtained from preliminary recall results based on the semantic vector via a two-stage accurate recall, i.e. wherein the processor is further configured to perform processes of: computing multiple similarities between a user input content of the input information and the multiple candidate texts of the text knowledge database;)
and obtaining the at least one of the multiple candidate texts based on the multiple similarities. See FIG. 4 & Paragraph [0067], (FIG. 4 illustrates the method comprising step S403 of obtaining candidate documents from a document library based on first and second candidate documents obtained having highest semantic similarities, i.e. obtaining the at least one of the multiple candidate texts based on the multiple similarities.)
Doggett, Gelfenbeyn and Cui are analogous art because they are in the same field of endeavor, conversational LLMs. It would have been obvious to anyone having ordinary skill in the art before the effective filing date to modify the system of Doggett-Gelfenbeyn to include the process of obtaining candidate documents based on similarity metrics as disclosed by Cui. Paragraph [0059] of Cui discloses that the use of a semantic vector describing each paragraph of a document may be used such that the accuracy of subsequent document recall and ranking may be improved.
Regarding independent claim 12,
The claim is analogous to the subject matter of dependent claim 2 directed to a method or process and is rejected under similar rationale.
Claim(s) 3, 8, 10, 13, 18 and 20 is/are rejected under 35 U.S.C. 103 as being unpatentable over Doggett in view of Gelfenbeyn as applied to claim 1 above, and further in view of Cui et al. (US PGPUB No. 2024/0411790; Pub. Date: Dec. 12, 2024).
Regarding dependent claim 3,
As discussed above with claim 1, Doggett-Gelfenbeyn discloses all of the limitations.
Doggett-Gelfenbeyn does not disclose the step wherein the processor is further configured to perform processes of: generating a prompt based on a user input content of the input information and the at least one of the multiple candidate texts;
and inputting the prompt to a large language model to generate the first output text.
Sun discloses the step wherein the processor is further configured to perform processes of: generating a prompt based on a user input content of the input information and the at least one of the multiple candidate texts; See Paragraphs [0118]-[0119], (Disclosing a chatbot system for an interactive platform. A user prompt 644 is illustrated as comprising natural language text such as "Do you have any dog food recommendations". Retrieving component 624 may identify a memory most relevant to the user prompt that is used to generate an augmented prompt, i.e. generating a prompt based on a user input content of the input information (e.g. the user natural language text) and the at least one of the multiple candidate texts (e.g. user memory data);)
and inputting the prompt to a large language model to generate the first output text. See Paragraph [0119], (The user's natural language prompt 644 is provided to the chatbot system which is used to determine a response and utilizes context-appropriate memories to generate an augmented prompt, i.e. inputting the prompt to a large language model to generate the first output text (e.g. the chatbot system determines an answer for a prompt as well as context-appropriate memories 636 to generated the augmented prompt used to generate the personalized response).)
Doggett, Gelfenbeyn and Sun are analogous art because they are in the same field of endeavor, conversational LLMs. It would have been obvious to anyone having ordinary skill in the art before the effective filing date to modify the system of Doggett-Gelfenbeyn to include the method of generating an augmented prompt as disclosed by Sun. Paragraphs [0023] & [0025] of Sun disclose that the chatbot provides a persistent memory that summarizes and stores memories from past conversations in a persistent datastore, wherein the stored data allows the model to leverage long-term user memory to improve responses.
Regarding dependent claim 8,
As discussed above with claim 1, Doggett-Gelfenbeyn discloses all of the limitations.
Doggett-Gelfenbeyn does not disclose the step wherein the text knowledge database comprises the multiple candidate texts, and multiple metadata and multiple vector information of the multiple candidate texts.
Sun discloses the step wherein the text knowledge database comprises the multiple candidate texts, and multiple metadata and multiple vector information of the multiple candidate texts. See Paragraph [0088], (Disclosing a chatbot system for an interactive platform. The knowledge base of the chatbot system 300 includes a set of information such as a predefined set of intents, entities, and responses, as well as external sources of information such as databases or APIs. Note [0080] wherein keywords and concepts are aggregated and mapped to a user as part of an intent profile or intent vector having weighted keywords and concepts based on the importance of said keywords/concepts in a conversation, i.e. wherein the text knowledge database comprises the multiple candidate texts (e.g. conversation history messages between users and the chatbot system), and multiple metadata (e.g. the multiple types of data associated with the knowledge base) and multiple vector information of the multiple candidate texts (e.g. intent vector data corresponds to user intents associated with user messages that form conversation histories between users and the chatbot system).
Doggett, Gelfenbeyn and Sun are analogous art because they are in the same field of endeavor, conversational LLMs. It would have been obvious to anyone having ordinary skill in the art before the effective filing date to modify the system of Doggett-Gelfenbeyn to include the method of generating an augmented prompt as disclosed by Sun. Paragraphs [0023] & [0025] of Sun disclose that the chatbot provides a persistent memory that summarizes and stores memories from past conversations in a persistent datastore, wherein the stored data allows the model to leverage long-term user memory to improve responses.
Regarding dependent claim 10,
As discussed above with claim 1, Doggett-Gelfenbeyn discloses all of the limitations.
Doggett-Gelfenbeyn does not disclose the step wherein the memory further stores a user database, wherein the user database comprises multiple basic user information corresponding to multiple users and multiple domain information.
Sun discloses the step wherein the memory further stores a user database, wherein the user database comprises multiple basic user information corresponding to multiple users and multiple domain information. See FIG. 3, (Disclosing a chatbot system for an interactive platform. FIG. 3 illustrates a system comprising interactive platform user data 324 comprising user profiles 314. The user profile data is used to determine an intent of a user. Note [0087] wherein user data of a user profile 314 may include conversation data information relating to a series of interactive sessions that the system may use to determine a context for conversations that occur over a plurality of sessions, i.e. wherein the memory further stores a user database, wherein the user database comprises multiple basic user information corresponding to multiple users and multiple domain information.)
Doggett, Gelfenbeyn and Sun are analogous art because they are in the same field of endeavor, conversational LLMs. It would have been obvious to anyone having ordinary skill in the art before the effective filing date to modify the system of Doggett-Gelfenbeyn to include the method of generating an augmented prompt as disclosed by Sun. Paragraphs [0023] & [0025] of Sun disclose that the chatbot provides a persistent memory that summarizes and stores memories from past conversations in a persistent datastore, wherein the stored data allows the model to leverage long-term user memory to improve responses.
Regarding dependent claim 13,
The claim is analogous to the subject matter of dependent claim 3 directed to a method or process and is rejected under similar rationale.
Regarding dependent claim 18,
The claim is analogous to the subject matter of dependent claim 8 directed to a method or process and is rejected under similar rationale.
Regarding dependent claim 20,
The claim is analogous to the subject matter of dependent claim 10 directed to a method or process and is rejected under similar rationale.
Claim(s) 4-5 and 14-15 is/are rejected under 35 U.S.C. 103 as being unpatentable over Doggett in view of Gelfenbeyn as applied to claim 1 above, and further in view of Sun et al. (US PGPUB No. 2024/0414108; Pub. Date: Dec. 12, 2024) and BHUPATI et al. (US PGPUB No. 2024/0378396; Pub. Date: Nov. 14, 2024)
Regarding dependent claim 4,
As discussed above with claim 1, Doggett-Gelfenbeyn discloses all of the limitations.
Doggett-Gelfenbeyn does not disclose the step wherein the processor is further configured to perform processes of: generating a prompt based on the input information and the character database;
Sun discloses the step wherein the processor is further configured to perform processes of: generating a prompt based on the input information and the character database; See Paragraph [0120], (FIG. 6B illustrates method 600 comprising step 612 wherein chatbot system 300 generates an augmented prompt 640 using context-appropriate memories 636 and user prompt 644, i.e. generating a prompt based on the input information (e.g. the user prompt) and the character database (e.g. the memory database data used to generate the augmented prompt);)
Doggett, Gelfenbeyn and Sun are analogous art because they are in the same field of endeavor, conversational LLMs. It would have been obvious to anyone having ordinary skill in the art before the effective filing date to modify the system of Doggett-Gelfenbeyn to include the method of generating an augmented prompt as disclosed by Sun. Paragraphs [0023] & [0025] of Sun disclose that the chatbot provides a persistent memory that summarizes and stores memories from past conversations in a persistent datastore, wherein the stored data allows the model to leverage long-term user memory to improve responses.
Doggett-Gelfenbeyn-Sun does not disclose the step of inputting the prompt to a large language model to obtain the first character template having highest confidence score.
BHUPATI discloses the step of inputting the prompt to a large language model to obtain the first character template having highest confidence score. See Paragraph [0024], (Disclosing a system for integrating LLM services for content analysis. The application service may comprise an LLM integration that generates a prompt upon determine the intent of a user or the purpose of a piece of content as the user is interacting with the application service. The system may generate a prompt including the content and context associate with the user based on determining a sufficiently high level of confidence or certainty.) See Paragraph [0048], (The prompt is configured according to a prompt template selected by LLM integration 121 based on the intent or purpose of the content, i.e. inputting the prompt to a large language model (e.g. a user input such as the user prompt 644 of Sun) to obtain the first character template having highest confidence score (e.g. confidence scores are used by the LLM to generate a prompt according to selected prompt templates).)
Doggett, Gelfenbeyn, Sun and BHUPATI are analogous art because they are in the same field of endeavor, conversational LLMs. It would have been obvious to anyone having ordinary skill in the art before the effective filing date to modify the system of Doggett-Gelfenbeyn-Sun to include the method of integrating LLM services for content analysis as disclosed by BHUPATI. Paragraph [0031] of BHUPATI discloses that the system may focus the generative activity of an LLM to improve its performance without exceeding token limits by balancing prompt size with providing sufficient information to generate a useful response. Reducing or optimizing the token count of the prompt reduces the costs associated with using the LLM and results in faster processing.
Regarding dependent claim 5,
As discussed above with claim 4, Doggett-Gelfenbeyn-Sun-BHUPATI discloses all of the limitations.
Sun further discloses the step wherein the input information comprises user input content, basic user information, and domain information. See Paragraph [0118], (A user prompt 644 is illustrated as comprising natural language text such as "Do you have any dog food recommendations", i.e. wherein the input information comprises user input content (e.g. the natural language text), basic user information (e.g. Note [0123] wherein user conversation history is maintained in a memories datastore 622 which stores interaction data using a user id as a key of a key-value pair), and domain information (e.g. the prompt is directed to a particular subject, in the example above the prompt relates to dog food).)
Regarding dependent claim 14,
The claim is analogous to the subject matter of dependent claim 4 directed to a method or process and is rejected under similar rationale.
Regarding dependent claim 15,
The claim is analogous to the subject matter of dependent claim 5 directed to a method or process and is rejected under similar rationale.
Claim(s) 6 and 16 is/are rejected under 35 U.S.C. 103 as being unpatentable over Doggett in view of Gelfenbeyn, Sun and BHUPATI as applied to claim 4 above, and further in view of Cui et al. (US PGPUB No. 2024/0411790; Pub. Date: Dec. 12, 2024).
Regarding dependent claim 6,
As discussed above with claim 4, Doggett-Gelfenbeyn-Sun-BHUPATI discloses all of the limitations.
Sun further discloses the step wherein multiple candidate dialogue examples of the multiple dialogue examples correspond to the first character template, wherein the processor is further configured to perform processes of: transferring a user input content of the input information into input content vector information; See Paragraph [0086], (Chatbot system 300 collects a set of prompts during an interactive session and maps the set of prompts to a set of keywords and/or concepts including an intent vector. The system may then determine the intent of the user based on the intent vector including weighted keywords and/or concepts, wherein the weights of keywords are determined based on an importance score calculated for each keyword and/or concept, i.e. transferring a user input content of the input information into input content vector information;)
Doggett-Gelfenbeyn-Sun-BHUPATI does not disclose the step of transferring the multiple candidate dialogue examples into multiple vector information;
and selecting one of the multiple candidate dialogue examples corresponding to a first vector information of the multiple vector information as the at least one of the multiple dialogue examples corresponding to the first character template when a similarity between the first vector information of the multiple vector information and the input content vector information is greater than a threshold.
Cui discloses the step of transferring the multiple candidate dialogue examples into multiple vector information; See FIG. 4 & Paragraph [0067], (Disclosing a system for answer information generation based on a large language model. FIG. 4 illustrates the method comprising step S403 of obtaining candidate documents from a document library based on first and second candidate documents obtained having highest semantic similarities. Note [0058] wherein a document semantic vector may be generated for each preset document in the document library, i.e. transferring the multiple candidate dialogue examples into multiple vector information (e.g. by generating vectors associated documents of a document library to which the process of FIG. 4 is applied, i.e. the candidate documents have associated document semantic vectors).)
and selecting one of the multiple candidate dialogue examples corresponding to a first vector information of the multiple vector information as the at least one of the multiple dialogue examples corresponding to the first character template when a similarity between the first vector information of the multiple vector information and the input content vector information is greater than a threshold. See Paragraphs [0066]-[0067], (Disclosing a system for answer information generation based on a large language model. The system may obtain candidate documents from a document library based on at least two of the semantic vector of the question text, the at least one piece of argument information, and the event category. Documents having the highest semantic similarities are obtained from preliminary recall results based on the semantic vector via a two-stage accurate recall, i.e. selecting one of the multiple candidate dialogue examples corresponding to a first vector information of the multiple vector information as the at least one of the multiple dialogue examples corresponding to the first character template when a similarity between the first vector information of the multiple vector information and the input content vector information is greater than a threshold (e.g. candidate documents are evaluated based on similarity metrics).)
Doggett, Gelfenbeyn, Sun, BHUPATI and Cui are analogous art because they are in the same field of endeavor, conversational LLMs. It would have been obvious to anyone having ordinary skill in the art before the effective filing date to modify the system of Doggett-Gelfenbeyn-Sun-BHUPATI to include the process of obtaining candidate documents based on similarity metrics as disclosed by Cui. Paragraph [0059] of Cui discloses that the use of a semantic vector describing each paragraph of a document may be used such that the accuracy of subsequent document recall and ranking may be improved.
Regarding dependent claim 16,
The claim is analogous to the subject matter of dependent claim 6 directed to a method or process and is rejected under similar rationale.
Claim(s) 7 and 17 is/are rejected under 35 U.S.C. 103 as being unpatentable over Doggett in view of Gelfenbeyn as applied to claim 1 above, and further in view of Catalano et al. (US PGPUB No. 2024/0291779; Pub. Date: Aug. 29, 2024) and Gadd (US PGPUB No. 2005/0125232; Pub. Date; Jun. 9, 2005)
Regarding dependent claim 7,
As discussed above with claim 1, Doggett-Gelfenbeyn discloses all of the limitations.
Gelfenbeyn further discloses the step wherein the processor is further configured to perform processes of: generating a prompt based on the input information, the first character template, the at least one of the multiple dialogue examples, and the first output text; See FIG. 5 & Paragraphs [0074]-[0075], (At Step C of method 500, the system may perform step 510 of composing intermediate outputs into templated formats for ingestion by large language models, animation, gesture and action systems. Step D then uses primary models and systems to generate final behavior-aligned data outputs, i.e. wherein the processor is further configured to perform processes of: generating a prompt based on the input information (e.g. the intermediate outputs are provided as inputs to LLMs) , the first character template (e.g. the primary models and systems used to generate the behavior-aligned data outputs), the at least one of the multiple dialogue examples, and the first output text (e.g. intermediate outputs generated at step 508 are fed into the templates);
The examiner notes that Gelfenbeyn discloses a multi-step method of receiving an input, generating an intermediate output, providing said intermediate output to a template and generating in-character outputs. While Gelfenbeyn does not disclose the use of "multiple dialogue examples" to generate an output, Doggett is relied upon to disclose the "multiple dialogue examples" of the claimed invention via the "original source vocabulary" concept described in at least Paragraphs [0046]-[0047] & [0070]. Therefore, one of ordinary skill in the art would recognize that the combination of Gelfenbeyn and Doggett would result in a template that may incorporate additional types of data when generating a response, such as by including elements of an "original source vocabulary" to ensure that a generated output speech is consistent with character parameters.
and inputting the prompt to a large language model to generate the second output text; See FIG. 5 & Paragraph [0078], (FIG. 5 illustrates method 500 comprising step D including step 514 of using primary models and systems to generate final behavior-aligned data outputs from the intermediate outputs that are ingested by large language models at step 510, i.e. inputting the prompt to a large language model to generate the second output text;)
and feeding the first output text to the answer field. See FIG. 5 & Paragraph [0073], (FIG. 5 comprises Step C of orchestrating which includes composing intermediate outputs, i.e. first output text, and providing the intermediate outputs into primary models as in step 512. The orchestration step comprises composing data from separate models into a specified templated format, i.e. and feeding the first output text to the answer field (e.g. the intermediate outputs are composed into a templated format for generating final behavior-aligned data outputs at step 514)
Doggett-Gelfenbeyn does not disclose the step wherein generating the prompt comprises:accessing a prompt template comprising a personal field, a history dialogue field, a query field, and an answer field;
and feeding the character description of the first character template to the personal field,
feeding a user input content of the input information to the query field,
Catalano discloses the step wherein generating the prompt comprises:accessing a prompt template comprising a personal field, a history dialogue field, a query field, and an answer field; See FIG. 9A & Paragraph [0160], (Disclosing a chatbot system for an interactive platform configured to generate responses in the form of interactive platform posts for group chats with multiple users or a single user. FIG. 9A illustrates method 900a wherein chatbot system 300 may generate a response using a received user prompt describing a desired persona of a chatbot. The chatbot system 300 generates a prompt for the generative AI model using the persona description and a prompt provided by a user, i.e. wherein generating the prompt comprises: accessing a prompt template comprising a personal field.)
The examiner notes that Catalano does not explicitly disclose the prompt template comprising "a history dialogue field, and an answer field;"
and feeding the character description of the first character template to the personal field, See FIG. 9A & Paragraph [0160], (FIG. 9A illustrates method 900a wherein chatbot system 300 may generate a response using a received user prompt describing a desired persona of a chatbot. The chatbot system 300 generates a prompt for the generative AI model using the persona description and a prompt provided by a user, i.e. feeding the character description of the first character template to the personal field (e.g. the prompt fed into AI model 334 uses the persona description).)
feeding a user input content of the input information to the query field, See FIG. 9A & Paragraph [0160], (FIG. 9A illustrates method 900a wherein chatbot system 300 may generate a response using a received user prompt describing a desired persona of a chatbot. The chatbot system 300 generates a prompt for the generative AI model using the persona description and a prompt provided by a user, i.e. feeding the character description of the first character template to the personal field (e.g. the prompt fed into AI model 334 uses the prompt provided by a user).)
Doggett, Gelfenbeyn and Catalano are analogous art because they are in the same field of endeavor, conversational applications. It would have been obvious to anyone having ordinary skill in the art before the effective filing date to modify the system of Doggett-Gelfenbeyn to include the process of generating chatbot outputs as disclosed by Catalano. Paragraph [0088] of Catalano discloses that chatbot system 300 may use user intent data and conversation context to improve its response and optimization capabilities over time by ingesting user feedback.
Doggett-Gelfenbeyn-Catalano does not disclose the step of feeding the at least one of the multiple dialogue example to the history dialogue field,
Gadd discloses the step of feeding the at least one of the multiple dialogue example to the history dialogue field, See Paragraph [0141], (Disclosing a system for creating and hosting speech-enabled applications having a speech interface that can be customized by a user. A speech application may be established using one or more templates which may include standard constructs such as prompts, grammars and dialogue flows, i.e. feeding the at least one of the multiple dialogue examples to the history dialogue field (e.g. the template comprising a dialogue flow attribute).)
Doggett-Gelfenbeyn-Catalano and Gadd are analogous art because they are in the same field of endeavor, conversational applications. It would have been obvious to anyone having ordinary skill in the art before the effective filing date to modify the system of Doggett-Gelfenbeyn-Catalano to include the process of generating speech applications based on one or more templates for customizing speech Gadd. Paragraphs [0014]-[0015] of Gadd disclose that The templates may be modified by AL algorithms such that the modification of templates is performed automatically, which represents an improvement whereby templates are updated without incurring significant system down-time.
Regarding dependent claim 17,
The claim is analogous to the subject matter of dependent claim 7 directed to a method or process and is rejected under similar rationale.
Claim(s) 9 and 19 is/are rejected under 35 U.S.C. 103 as being unpatentable over Doggett in view of Gelfenbeyn as applied to claim 1 above, and further in view of Catalano et al. (US PGPUB No. 2024/0291779; Pub. Date: Aug. 29, 2024)
Regarding dependent claim 9,
As discussed above with claim 1, Doggett-Gelfenbeyn discloses all of the limitations.
Doggett-Gelfenbeyn does not disclose the step wherein the character database further comprises multiple text description information and multiple graphic description information corresponding to the multiple character templates.
Catalano discloses the step wherein the character database further comprises multiple text description information and multiple graphic description information corresponding to the multiple character templates. See Paragraph [0096], (Chatbot system 300 determines a personality or tone for responses generated by the chatbot system based on storing a conversation stage for a user as part of a user's profile. The chatbot system uses the conversation state and demographic or other data about the user to determine a personality or tone for the user. Note [0156] wherein the chatbot system 300 may allow a user to generate a persona description 910 comprising a textual description of a chatbot in order to generate a personality type that will be used to process user prompts, i.e. multiple graphic description information corresponding to the multiple character templates (e.g. a user may provide any description of a chatbot including describing characteristics. Note [0158] wherein a user may even assign a bitmoji or thumbnail image for the personal chatbot).
Doggett, Gelfenbeyn and Catalano are analogous art because they are in the same field of endeavor, chatbot systems. It would have been obvious to anyone having ordinary skill in the art before the effective filing date to modify the system of Doggett-Gelfenbeyn to include the method of customizing the tone and style of a chatbot's responses as disclosed by Catalano. Paragraph [0099] of Catalano discloses that the system may analyze a user's current activity, language patterns and emotional tone in order to craft prompts and responses that are appropriate to the user's emotional state. This context-aware approach allows the chatbot system 300 to provide a more intuitive and responsive experience.
Regarding dependent claim 19,
The claim is analogous to the subject matter of dependent claim 9 directed to a method or process and is rejected under similar rationale.
Examiner’s Input
The following references were considered relevant to the claimed invention but were not relied upon for the current prior art rejections:
Doggett et al. (US PGPUB No. 2024/0394965 A1; Pub. Date: Nov. 29, 2024)
Doggett’965 is directed to a system for synthesizing an interaction with a virtual character. The technique includes matching a first message from a user to a first set of memories associated with the virtual character and determining at least a portion of the first set of memories based on a plurality of factors associated with the first set of memories. The technique also includes inputting a first prompt that includes (i) one or more instructions associated with the virtual character, (ii) the at least a portion of the first set of memories, and (iii) the first message into a machine learning model (see Abstract) Paragraphs [0045], [0046] describe a process of generating memories 310 associated with virtual characters which may be stored in a database for subsequent retrieval and use.
Response to Arguments
Applicant’s arguments, see Notice of Appeal, filed 5/25/2026, with respect to the rejection(s) of claim(s) 1, 3, 7-11, 13 and 17-20 under 35 USC 103 have been fully considered and are persuasive. Therefore, the rejection has been withdrawn.
However, upon further consideration, a new ground(s) of rejection is made as discussed in the rejections above.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to Fernando M Mari whose telephone number is (571)272-2498. The examiner can normally be reached Monday-Friday 7am-4pm.
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, Ann J. Lo can be reached at (571) 272-9767. 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.
/FMMV/Examiner, Art Unit 2159
/ANN J LO/Supervisory Patent Examiner, Art Unit 2159
Prosecution Insights