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 .
Response to Amendment
This communication is responsive to the applicant’s amendment dated 12/19/2025. The applicant amended claims 1, 9, 12, 14-15, and 18. The applicant cancelled claims 13, 16-17, and 20. Lastly, the applicant added new claim 21.
Response to Arguments
Applicant’s arguments, see Remarks (pg. 8, line 14 – pg. 16, line 15), filed 12/19/2025, with respect to claims 1-20 have been fully considered and are persuasive. The 35 U.S.C. 101 rejection of claims 1-20 has been withdrawn.
Applicant's arguments with respect to 35 U.S.C. 103 filed 12/19/2025 have been fully considered but they are not persuasive.
The applicant alleges that Reddy fails to teach or suggest "an unstructured data manager that: retrieves results for the user query by executing an automatically-generated structured query on the one or more relational databases, wherein the automatically-generated structured query is generated by a currently selected Large Language Model (LLM)”. The applicant’s reasoning to support this argument can be found on pg. 17, line 21-27. However, in this section, the applicant does not argue how Reddy is different than what is being claimed. Instead, the applicant is merely summarizing the citations from Reddy that the examiner believes teaches this limitation. Nevertheless, the examiner has reviewed the limitation and believes it is being taught by Reddy. Additionally, the examiner has provided additional citations to support his position as well as addressed the amendments in the limitations below. Lastly, the applicant states (pg. 19, line 12) that one of ordinary skill in the art would have found no motivation to combine Garcia, Reddy, Zhao, and Eberlein. The examiner respectfully disagrees given they are all in the same field of endeavor. Therefore, the 35 U.S.C. 103 rejection is maintained.
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.
Claims 1-3, 5-10, 12, 14-15, 18-19, and 21 are rejected under 35 U.S.C. 103 as being unpatentable over Garcia et al. US 12211502 B2 (hereinafter Garcia) in view of Reddy et al. US 20250086394 A1 (hereinafter Reddy) in view of Eberlein et al. US 20250156574 A1 (hereinafter Eberlein).
Regarding independent claim 1, Garcia teaches a Generative Artificial Intelligence (AI) based chatbot apparatus, comprising:
at least one hardware processor (FIG. 2A, 220); and
at least one non-transitory processor-readable medium storing instructions for and the at least one hardware processor executing (FIG. 6B, 618, ([Column 28, line 65 – Column 29, line 1]“Memory 618 of personal electronic device 600 is a non-transitory computer-readable storage medium, for storing computer-executable instructions, which, when executed by one or more computer processors 616”)):
a chatbot interface that: receives a user query in a natural language, the user query includes an intent indicative of a task to be executed and a slot indicative of relevant entities from a plurality of data sources to be used for the task execution, wherein the plurality of data sources include one or more relational databases and one or more unstructured knowledge bases (FIG. 9, 912, [Column 6, line 25-29 “One or more processing modules 114 utilize data and models 116 to process speech input and determine the user's intent based on natural language input. Further, one or more processing modules 114 perform task execution based on inferred user intent.”]; FIG. 7C, [Column 38, line 1-21] “In the present discussion, each domain is associated with a respective actionable intent…”, examiner interprets slots as domains; FIG. 1, 116, 120, [Column 6, line 24-33] “One or more processing modules 114 utilize data and models 116 to process speech input and determine the user's intent based on natural language input. Further, one or more processing modules 114 perform task execution based on inferred user intent. In some examples, DA server 106 communicates with external services 120 through network(s) 110 for task completion or information acquisition. I/O interface to external services 118 facilitates such communications”, examiner interprets 116 and 120 as databases);
an intent-slot prediction model that: outputs predictions for the intent and the slot based on the user query ([Column 4, line 56 – Column 4, line 59] “The techniques also include performing candidate intent evaluation without actual execution (e.g., making a dry run), thereby determining whether a candidate intent is actionable”, examiner interprets intent evaluation as predictions); and
outputs confidences associated with the predictions; an orchestrator that: determines if the confidence associated with the intent prediction is above or below a configured confidence limit ([Column 39, line 47 – Column 39, line 51] “Based on the candidate text representation and the associated contextual information, the one or more machine learning mechanism are configured to determine intent confidence scores over a set of candidate actionable intents”; [Column 50, line 47 – Column 50, line 50] “CIE 860 can estimate a confidence level associated with performing a task and determine whether the confidence level associated with performing the task satisfies a threshold confidence level”);
a structured data manager that: retrieves results for the user query from the one or more relational databases with a structured query mapped to the intent if the prediction for the intent by intent-slot prediction model has a higher accuracy than the configured confidence limit ([Column 39, line 63-67] “once natural language processing module 732 identifies an actionable intent (or domain) based on the user request, natural language processing module 732 generates a structured query to represent the identified actionable intent.”; [Column 40, line 34-37] “a respective structured query (partial or complete) is generated for each identified candidate actionable intent. Natural language processing module 732 determines an intent confidence score for each candidate actionable intent and ranks the candidate actionable intents based on the intent confidence scores”);
Garcia fails to teach an intent-slot prediction model trained with training data including various forms of user queries labeled with corresponding intents and slots, the intent-slot prediction model; an unstructured data manager that: retrieves results for the user query by executing an automatically-generated structured query on the one or more relational databases, if the prediction for the intent has a lower accuracy than the configured confidence limit, wherein the automatically-generated structured query is generated by a currently selected Large Language Model (LLM); and a Generative AI switch (Gen. AI) that: switches an LLM from a plurality of LLMs based on the user query and user feedback, wherein each LLM of the plurality is implemented using a different algorithm, and wherein the Gen. AI switch further applies a rule-based procedure to deselect the currently selected LLM and select another LLM from the plurality of LLMs based on the user query and the user feedback.
However, Reddy teaches an intent-slot prediction model trained with training data including various forms of user queries labeled with corresponding intents and slots, the intent-slot prediction model ([0261] “training the one or more large language models with examples of invocation logic definitions associated with respective intents”);
an unstructured data manager that: retrieves results for the user query by executing an automatically-generated structured query on the one or more relational databases, if the prediction for the intent has a lower accuracy than the configured confidence limit, wherein the automatically-generated structured query is generated by a currently selected Large Language Model (LLM) ([0091] “The response may be in a format (e.g., JSON) from which the relevant information is extracted, interpreted, or the like, and then incorporated into a response”, examiner interprets JSON as the structured query; [0078-0079] “Because the large language model is loaded with documents that provide appropriate context for the digital assistant domain at issue, the large language model tailors outputs to the domain; [0020], [0028], FIG. 9, 940, [0084] “Various techniques can be used to select the top intents. For example, the top n intents can then be selected, those intents exceeding a threshold priority score can be selected, or the like”, the examiner interprets the threshold as the confidence limit).
wherein the Gen. AI switch further applies a rule-based procedure to deselect the currently selected LLM and select another LLM from the plurality of LLMs based on the user query and the user feedback ([0154] 5. The “Building a Bot” section contains step-by-step tutorials that walk you through the process of creating a chatbot, starting from creating intents to deploying your bot on various channels;)
Garcia in view of Reddy are considered to be analogous to the claimed invention because both are the same field of providing natural language interaction by virtual assistants. 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 the techniques providing natural language interaction by virtual assistants of Garcia with the technique of using LLM to generate structured queries taught by Reddy in order to improve automated generation of digital assistants with large language models (see Reddy [0001]).
Garcia in view of Reddy fails to teach a Generative AI switch (Gen. AI) that: switches an LLM from a plurality of LLMs based on the user query and user feedback, wherein each LLM of the plurality is implemented using a different algorithm
However, Eberlein teaches a Generative AI switch (Gen. AI) that: switches an LLM from a plurality of LLMs based on the user query and user feedback, wherein each LLM of the plurality is implemented using a different algorithm ([0056] “The access log can be processed with user feedback on the quality of the application with respect to particular queries, to optimize the application and create a newer version, with an adjusted (changed) configuration indicative of a selection of LLM and data access”).
Garcia in view of Reddy in view of Eberlein are considered to be analogous to the claimed invention because both are the same field of providing natural language interaction by virtual assistants. 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 the techniques providing natural language interaction by virtual assistants of Garcia in view of Reddy with the technique of selecting LLMs based on user feedback taught by Eberlein in order to improve data access control for large language model services. (see Eberlein [0001])
Regarding claim 2, Garcia in view of Reddy in view of Eberlein teaches all of the limitations of claim 1, upon which claim 2 depends.
Additionally, Garcia teaches wherein the chatbot interface executed by the at least one hardware processor further: formats the retrieved results as an answer to the user query; and
outputs the answer to a user device (FIG. 13B).
Regarding claim 3, Garcia in view of Reddy in view of Eberlein teaches all of the limitations of claim 1, upon which claim 3 depends.
Additionally, Garcia teaches wherein the chatbot interface executed by the at least one hardware processor further: formats the answer as one or more of a textual response and, an audio response ([Column 6, 4-6] “the digital assistant also provides responses in other visual or audio forms, e.g., as text, alerts, music, videos, animations, etc.”).
Regarding claim 5, Garcia in view of Reddy in view of Eberlein teaches all of the limitations of claim 1, upon which claim 5 depends.
Additionally, Garcia teaches wherein the orchestrator executed by the at least one hardware processor further: forwards, based on the slot, the user query to one of the structured data manager and the unstructured data manager if the confidence associated with the intent prediction is above the configured confidence limit; and forwards the user query to the unstructured data manager if the confidence associated with the intent prediction is below the configured confidence limit ([Column 40, line 34-45] “Natural language processing module 732 determines an intent confidence score for each candidate actionable intent and ranks the candidate actionable intents based on the intent confidence scores. In some examples, natural language processing module 732 passes the generated structured query (or queries), including any completed parameters, to task flow processing module 736 (“task flow processor”). In some examples, the structured query (or queries) for the m-best (e.g., m highest ranked) candidate actionable intents are provided to task flow processing module 736, where m is a predetermined integer greater than zero”).
Regarding claim 6, Garcia in view of Reddy in view of Eberlein teaches all of the limitations of claim 1, upon which claim 6 depends.
Additionally, Garcia teaches wherein the structured data manager executed by the at least one hardware processor retrieves the results from the one or more relational databases by: determining that the accuracy of the intent associated with the user query is above the configured confidence limit ([Column 63, line 21-30] “At block 1584, whether the confidence level associated with performing the task satisfies a threshold confidence level is determined. At block 1586, in accordance with a determination that the confidence level associated with performing the task satisfies the threshold confidence level, it is determined that the task can be performed”).
Regarding claim 7, Garcia in view of Reddy in view of Eberlein teaches all of the limitations of claim 6, upon which claim 7 depends.
Additionally, Garcia teaches wherein the structured data manager executed by the at least one hardware processor retrieves the results from the one or more relational databases by: selecting the structured query from preconfigured mappings based on the intent ([Column 40, line 8-10] “According to the ontology, a structured query for a “restaurant reservation” domain includes…” FIG. 10, 840, [Column 48, line 34-37] “using a decision tree, the FTM of the virtual assistant can estimate a likelihood that the utterance corresponding to the first candidate text representation is not directed to the virtual assistant”;).
Regarding claim 8, Garcia in view of Reddy in view of Eberlein teaches all of the limitations of claim 1, upon which claim 8 depends.
Additionally, Reddy teaches wherein the unstructured data manager executed by the at least one hardware processor retrieves the results from the plurality of data sources by: determining that a prompt can be automatically generated from the user query, wherein the prompt is provided to the currently selected LLM for automatic generation of the structured query corresponding to the intent specified in the user query ([0049] “Prompting can also be performed to identify entities. For example, at least one of the one or more large language models can be prompted to identify one or more entities in the utterances”; [0067] The input documents can be domain-specific documents that help the large language model generate skills, intents, and entities appropriate to the domain).
Regarding claim 9, Garcia in view of Reddy in view of Eberlein teaches all of the limitations of claim 8, upon which claim 9 depends.
Additionally, Reddy teaches wherein if it is determined that the prompt can be automatically generated from the user query the unstructured data manager executed by the at least one hardware processor retrieves the results from the plurality of data sources by: identifying within the prompt, a context including a user profile associated with the user query and a selection of the currently selected LLM from the plurality of LLMs ([0068] “intents can be generated using large language models. In the context of digital assistant development, an intent represents the purpose or goal that a user wants to achieve through their interaction with the digital assistant.” [0078] “Because the large language model is loaded with documents that provide appropriate context for the digital assistant domain at issue, the large language model tailors outputs to the domain”; [0152] 3. Clicking on “Getting Started” will take you to a page that explains how to create an account and start building your first bot ).
Regarding claim 10, Garcia in view of Reddy in view of Eberlein teaches all of the limitations of claim 8, upon which claim 10 depends.
Additionally, Reddy teaches wherein if it is determined that the prompt cannot be automatically generated from the user query the unstructured data manager executed by the at least one hardware processor retrieves the results from the plurality of data sources by: generating a sequence of sub-tasks for the task conveyed in the intent ([0032] “the orchestrator 110 can perform the methods described herein. Agents can be employed by the orchestrator 110 to perform various subtasks of the automated development process”);
and automatically generating a sequence of structured queries corresponding to the sequence of sub-tasks ([0091] The response may be in a format (e.g., JSON) from which the relevant information is extracted, interpreted, or the like, and then incorporated into a response.).
Regarding claim 12, Garcia in view of Reddy in view of Eberlein teaches all of the limitations of claim 1, upon which claim 12 depends.
Additionally, Reddy teaches wherein the at least one non-transitory processor-readable medium storing instructions for and the at least one hardware processor further executing:
a feedback processor that: receives the user feedback regarding the retrieved results provided in response to the user query ([0168] 6. View bot analytics—this intent can be used to view metrics and analytics for a bot, such as the number of users, messages, and user satisfaction).
Regarding claim 14, Garcia in view of Reddy in view of Eberlein teaches all of the limitations of claim 1, upon which claim 14 depends.
Additionally, Reddy teaches wherein the plurality of LLMs include subsets of customized LLMs wherein each subset of customized LLMs is customized for a given domain, wherein each LLM within the set of customized LLMs is further trained to identify a corresponding intent ([0188-0208] examples of customizable bots that use a domain specific LLM; [0067] The input documents can be domain-specific documents that help the large language model generate skills, intents, and entities appropriate to the domain. Thus, in any of the examples herein, a domain-specific digital assistant can be generated via domain-specific input documents).
Regarding independent claim 15, Garcia teaches A Generative Artificial Intelligence (Gen. AI) based data retrieval method, comprising:
receiving a user query in a natural language, the user query includes an intent indicative of a task to be executed and a slot indicative of one or more relevant entities from a plurality of data sources to be used for the task execution, wherein the plurality of data sources include one or more relational databases and one or more knowledge bases storing unstructured data (FIG. 9, 912, [Column 6, line 25-29 “One or more processing modules 114 utilize data and models 116 to process speech input and determine the user's intent based on natural language input. Further, one or more processing modules 114 perform task execution based on inferred user intent.”]; FIG. 7C, [Column 38, line 1-21] “In the present discussion, each domain is associated with a respective actionable intent…”, examiner interprets slots as domains; FIG. 1, 116, 120, [Column 6, line 24-33] “One or more processing modules 114 utilize data and models 116 to process speech input and determine the user's intent based on natural language input. Further, one or more processing modules 114 perform task execution based on inferred user intent. In some examples, DA server 106 communicates with external services 120 through network(s) 110 for task completion or information acquisition. I/O interface to external services 118 facilitates such communications”, examiner interprets 116 and 120 as databases);
extracting predictions for the intent and the slot based on the user query along with confidences associated with the predictions ([Column 4, line 56 – Column 4, line 59] “The techniques also include performing candidate intent evaluation without actual execution (e.g., making a dry run), thereby determining whether a candidate intent is actionable”, examiner interprets intent evaluation as predictions; [Column 39, line 47 – Column 39, line 51] “Based on the candidate text representation and the associated contextual information, the one or more machine learning mechanism are configured to determine intent confidence scores over a set of candidate actionable intents”);
determining that the confidence associated with the intent prediction is below a configured confidence limit ([Column 50, line 47 – Column 50, line 50] “CIE 860 can estimate a confidence level associated with performing a task and determine whether the confidence level associated with performing the task satisfies a threshold confidence level”);
Garcia fails to teach wherein extracting the predictions for the intent and the slot by training an intent- slot prediction model with training data including various forms of user queries labeled with corresponding intents and slots; retrieving results for the user query from the one or more of relational databases and the knowledge bases with an automatically generated structured query, wherein the automatically generated structured query is generated by a currently selected Large Language Model (LLM) from a plurality of LLMs; wherein the switching comprises implementing a rule-based procedure to deselect the currently selected LLM based on the user feedback, and select another LLM from the plurality of LLMs;
However, Reddy teaches wherein extracting the predictions for the intent and the slot by training an intent- slot prediction model with training data including various forms of user queries labeled with corresponding intents and slots ([0261] “training the one or more large language models with examples of invocation logic definitions associated with respective intents”);
retrieving results for the user query from the one or more of relational databases and the knowledge bases with an automatically generated structured query, wherein the automatically generated structured query is generated by a currently selected Large Language Model (LLM) from a plurality of LLMs ([0091] “The response may be in a format (e.g., JSON) from which the relevant information is extracted, interpreted, or the like, and then incorporated into a response”, examiner interprets JSON as the structured query; [0078-0079] “Because the large language model is loaded with documents that provide appropriate context for the digital assistant domain at issue, the large language model tailors outputs to the domain”);
wherein the switching comprises implementing a rule-based procedure to deselect the currently selected LLM based on the user feedback, and select another LLM from the plurality of LLMs ([0154] 5. The “Building a Bot” section contains step-by-step tutorials that walk you through the process of creating a chatbot, starting from creating intents to deploying your bot on various channels);
Garcia in view of Reddy are considered to be analogous to the claimed invention because both are the same field of providing natural language interaction by virtual assistants. 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 the techniques providing natural language interaction by virtual assistants of Garcia with the technique of using LLM to generate structured queries taught by Reddy in order to improve automated generation of digital assistants with large language models (see Reddy [0001]).
Garcia in view of Reddy fails to teach switching based on the user query and user feedback, the currently selected LLM by selecting another LLM from the plurality of LLMs, wherein each of the plurality of LLMs is based on a different algorithm.
However, Eberlein teaches switching based on the user query and user feedback, the currently selected LLM by selecting another LLM from the plurality of LLMs, wherein each of the plurality of LLMs is based on a different algorithm ([0056] “The access log can be processed with user feedback on the quality of the application with respect to particular queries, to optimize the application and create a newer version, with an adjusted (changed) configuration indicative of a selection of LLM and data access”).
Garcia in view of Reddy in view of Eberlein are considered to be analogous to the claimed invention because both are the same field of providing natural language interaction by virtual assistants. 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 the techniques providing natural language interaction by virtual assistants of Garcia in view of Reddy with the technique of selecting LLMs based on user feedback taught by Eberlein in order to improve data access control for large language model services. (see Eberlein [0001]).
Regarding independent claim 18, Garcia teaches A non-transitory processor-readable storage medium comprising machine-readable instructions that cause a processor to: receive a user query in a natural language, the user query includes an intent indicative of a task to be executed and a slot indicative of a data source from a plurality of data sources to be used for the task execution, wherein the plurality of data sources include one or more relational databases and one or more knowledge bases storing unstructured data (FIG. 9, 912, [Column 6, line 25-29 “One or more processing modules 114 utilize data and models 116 to process speech input and determine the user's intent based on natural language input. Further, one or more processing modules 114 perform task execution based on inferred user intent.”]; FIG. 7C, [Column 38, line 1-21] “In the present discussion, each domain is associated with a respective actionable intent…”, examiner interprets slots as domains; FIG. 1, 116, 120, [Column 6, line 24-33] “One or more processing modules 114 utilize data and models 116 to process speech input and determine the user's intent based on natural language input. Further, one or more processing modules 114 perform task execution based on inferred user intent. In some examples, DA server 106 communicates with external services 120 through network(s) 110 for task completion or information acquisition. I/O interface to external services 118 facilitates such communications”, examiner interprets 116 and 120 as databases);
obtain predictions for the intent and identification of the slot based on the user query and confidences associated with the predictions ([Column 4, line 56 – Column 4, line 59] “The techniques also include performing candidate intent evaluation without actual execution (e.g., making a dry run), thereby determining whether a candidate intent is actionable”, examiner interprets intent evaluation as predictions);
determine if the confidence associated with the intent prediction is above or below a configured confidence limit ([Column 39, line 47 – Column 39, line 51] “Based on the candidate text representation and the associated contextual information, the one or more machine learning mechanism are configured to determine intent confidence scores over a set of candidate actionable intents”);
retrieve results for the user query from the one or more relational databases with a structured query mapped to the intent if the prediction for the intent has a higher accuracy than the configured confidence limit ([Column 39, line 63-67] “once natural language processing module 732 identifies an actionable intent (or domain) based on the user request, natural language processing module 732 generates a structured query to represent the identified actionable intent.”; [Column 40, line 34-37] “a respective structured query (partial or complete) is generated for each identified candidate actionable intent. Natural language processing module 732 determines an intent confidence score for each candidate actionable intent and ranks the candidate actionable intents based on the intent confidence scores”); and
Garcia fails to teach train the intent-slot prediction model with training data including various forms of user queries labeled with corresponding intents and slots; retrieve results for the user query by executing a structured query on the one or more relational databases if the prediction for the intent has a lower accuracy than the configured confidence limit, wherein the structured query is automatically generated by a currently selected Large Language Model (LLM); switch an LLM from a plurality of LLMs based on the user query and user feedback, wherein each LLM of the plurality is implemented using a different algorithm, and wherein the Gen. AI switch further applies a rule-based procedure to deselect the currently selected LLM and select another LLM from the plurality of LLMs based on the user query and the user feedback.
However, Reddy teaches train the intent-slot prediction model with training data including various forms of user queries labeled with corresponding intents and slots ([0261] “training the one or more large language models with examples of invocation logic definitions associated with respective intents”)
retrieve results for the user query by executing a structured query on the one or more relational databases if the prediction for the intent has a lower accuracy than the configured confidence limit, wherein the structured query is automatically generated by a currently selected Large Language Model (LLM) ([0091] “The response may be in a format (e.g., JSON) from which the relevant information is extracted, interpreted, or the like, and then incorporated into a response”, examiner interprets JSON as the structured query; [0078-0079] “Because the large language model is loaded with documents that provide appropriate context for the digital assistant domain at issue, the large language model tailors outputs to the domain.)
wherein the Gen. AI switch further applies a rule-based procedure to deselect the currently selected LLM and select another LLM from the plurality of LLMs based on the user query and the user feedback ([0154] 5. The “Building a Bot” section contains step-by-step tutorials that walk you through the process of creating a chatbot, starting from creating intents to deploying your bot on various channels;)
Garcia in view of Reddy are considered to be analogous to the claimed invention because both are the same field of providing natural language interaction by virtual assistants. 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 the techniques providing natural language interaction by virtual assistants of Garcia with the technique of using LLM to generate structured queries taught by Reddy in order to improve automated generation of digital assistants with large language models (see Reddy [0001]).
Garcia in view of Reddy fails to teach switch an LLM from a plurality of LLMs based on the user query and user feedback, wherein each LLM of the plurality is implemented using a different algorithm
However, Eberlein teaches switch an LLM from a plurality of LLMs based on the user query and user feedback, wherein each LLM of the plurality is implemented using a different algorithm ([0056] “The access log can be processed with user feedback on the quality of the application with respect to particular queries, to optimize the application and create a newer version, with an adjusted (changed) configuration indicative of a selection of LLM and data access”).
Garcia in view of Reddy in view of Eberlein are considered to be analogous to the claimed invention because both are the same field of providing natural language interaction by virtual assistants. 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 the techniques providing natural language interaction by virtual assistants of Garcia in view of Reddy with the technique of selecting LLMs based on user feedback taught by Eberlein in order to improve data access control for large language model services. (see Eberlein [0001])
Regarding claim 19, Garcia in view of Reddy in view of Eberlein teaches all of the limitations of claim 18, upon which claim 19 depends.
Additionally, Garcia teaches including further instructions that cause the processor to: format the retrieved results as an answer to the user query; and output the answer via a chatbot output interface (FIG. 13B).
Regarding claim 21, Garcia in view of Reddy in view of Eberlein teaches all of the limitations of claim 1, upon which claim 21 depends.
Additionally, Reddy teaches wherein the plurality of LLMs are trained for generating structured queries ([0026] “Although a single large language model 130 is shown as generating intents, in practice, one or more large language models 130 can be used by providing one or more documents to respective of the models 130”; [0091] The response may be in a format (e.g., JSON) from which the relevant information is extracted)
Claim 4, 11 are rejected under 35 U.S.C. 103 as being unpatentable over Garcia in view of Reddy in view of Eberlein as shown above in claim 1, in further view of Zhao et al. US 20250077511 A1 (hereinafter Zhao).
Regarding claim 4, Garcia in view of Reddy in view of Eberlein teaches all of the limitations of claim 1, upon which claim 4 depends.
Additionally, Garcia teaches further comprises the plurality of data sources wherein the one or more knowledge bases include at least user profile information ([Column 36, line 22 – Column 36, line 25] “the rank of the candidate pronunciation is based on one or more characteristics (e.g., geographic origin, nationality, ethnicity, etc.) of the user stored in the user's profile on the device”) and
Garcia in view of Reddy in view of Eberlein fails to teach database schema of the one or more relational databases
However, Zhao teaches database schema of the one or more relational databases ([0029] “The prompt generator may provide the natural language query and context information about the database, such as database schema information, database rules, database field descriptions, and information about vulnerabilities and/or alerts, as input to the generative model with an instruction to generate a database query (e.g., a SQL query) corresponding to the natural language query”)
Garcia in view of Reddy in view of Eberlein in view of Zhao are considered to be analogous to the claimed invention because both are the same field of providing natural language interaction by virtual assistants. 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 the techniques providing natural language interaction by virtual assistants of Garcia in view of Reddy in view of Eberlein with the technique of using database schema as knowledge bases taught by Zhao in order to improve a stateful chatbot system leverages generative AI to provide an interface by which users can retrieve information from backend IoT databases of a security provider via natural language queries (see Zhao [Abstract]).
Regarding claim 11, Garcia in view of Reddy in view of Eberlein teaches all of the limitations of claim 10, upon which claim 11 depends.
Garcia in view of Reddy in view of Eberlein fails to teach executing a search of the knowledge bases with the user query; and selecting top n documents that exceed a predetermined similarity threshold with the user query, wherein n is a natural number greater than or equal to 1; and obtaining the results from the currently-selected LLM by providing an input of the top n documents concatenated with the user query.
However, Zhao teaches executing a search of the knowledge bases with the user query ([0015]“The database query 106 may be a SQL query that searches the database 115 for the IoT devices documented…”); and
selecting top n documents that exceed a predetermined similarity threshold with the user query, wherein n is a natural number greater than or equal to 1 ([0015] “The database query 106 may be a SQL query that searches the database 115 for the IoT devices documented therein that are indicated to be a camera and have a risk score that exceeds a threshold corresponding to a higher severity of risk”); and
obtaining the results from the currently-selected LLM by providing an input of the top n documents concatenated with the user query ([0015] “[0016] The chatbot 101 queries the database 115 with the database query 106 and obtains results 108”).
Garcia in view of Reddy in view of Eberlein in view of Zhao are considered to be analogous to the claimed invention because both are the same field of providing natural language interaction by virtual assistants. 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 the techniques providing natural language interaction by virtual assistants of Garcia in view of Reddy in view of Eberlein with the technique of executing searches of databases of the top documents taught by Zhao in order to improve a stateful chatbot system leverages generative AI to provide an interface by which users can retrieve information from backend IoT databases of a security provider via natural language queries (see Zhao [Abstract]).
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Chen et al. (US 20250077792 A1) teaches technologies are capable of a training pipeline to fine-tune a machine learning model given a limited set of domain-specific data. The embodiments describe using a first machine learning model to generate a pseudo label associated with a domain-specific training document. The pseudo label comprises a machine-generated text of a content type extracted from the domain-specific training document. The embodiments further describe fine-tuning a second machine learning model using the pseudo label, the domain-specific training document, a first low-rank weight matrix, and a second low-rank weight matrix. The fine-tuned second machine learning model generates text of the content type from a domain-specific document.
Kumar et al. (US 11855860 B1) teaches a plurality of resolved incident tickets may each include a worklog providing a history of actions taken during attempts to resolve a corresponding resolved incident and a resolution having at least one resolution statement. An iterative processing of the plurality of resolved incident tickets may include processing each resolution statement of the resolution with at least one domain-specific statement classifier specific to the incident domain to either discard or retain a classified resolution statement; processing each retained classified resolution statement in conjunction with the worklog to determine whether to discard or retain the resolved incident; providing an updated resolution for the resolved incident when the resolved incident is retained, and adding the resolved incident with the updated resolution to the processed incident tickets. Then, at least one machine learning model may be trained to process a new incident ticket, using the processed incident tickets.
Krabach et al. (US 20250165714 A1) teaches a system is provided for managing specialized tasks and information retrieval processes. Agents are configured to perform tasks and/or retrieve information in a specialized domain. The system receives, via an interaction interface, a message from a user for the trained generative model to generate an output, generates a context of the message, generates a request including the context and the message, executes an orchestrator configured to: receive the request, determine, using semantic decision making, one or more agents to handle the request, input the request into one or more agents to perform a task and/or retrieve information in specialized domains, generate a prompt based on the retrieved information and/or the performed task and the message from the user, provide the prompt to the trained generative model, receive, in response to the prompt, a response from the trained generative model, and output the response to the user.
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to ZEESHAN SHAIKH whose telephone number is (703)756-1730. The examiner can normally be reached Monday-Friday 7:30AM-5:00PM.
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, Richemond Dorvil can be reached at (571) 272-7602. 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.
/ZEESHAN MAHMOOD SHAIKH/Examiner, Art Unit 2658
/RICHEMOND DORVIL/Supervisory Patent Examiner, Art Unit 2658