DETAILED ACTION
1. This action is responsive to Application no.19/004,997 filed 12/30/2024. All claims have been examined and are currently pending.
Notice of Pre-AIA or AIA Status
2. The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Information Disclosure Statement
3. The information disclosure statement (IDS) submitted is in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statement is being considered by the examiner.
Claim Rejections - 35 USC § 103
4. 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 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.
5. 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.
6. Claims 1-5, 7-13, 15-18, 20 are rejected under 35 U.S.C. 103 as being unpatentable over Thomas (12,242,817) in view of Hamilton et al (2024/0378223).
Regarding claim 1 Thomas (12,242,817) teaches A method (col 3 l 8-9 computer-implemented method; l 22-26: a system including one or more processors and one or more non-transitory computer-readable media storing computing instructions that, when executed on the one or more processors, cause the one or more processors to perform various operations) comprising:
based on obtaining a user query comprising natural language that corresponds to a first conversation session, updating a context of the first conversation session based on the user query, wherein updating the context of the first conversation session comprises extracting and inferring information from the user query and updating the context with the information (abstract; col 3 l 10-21 query is part of the conversation; context from the chat history; col 15 l. 3-21 “Tell me more about the last one”; col 17 l. 58-61 chat history, kept in cache memory);
rephrasing the user query based on the context of the first conversation session to generate a rephrased query, wherein rephrasing the user query comprises prompting a first language model with the user query and the context of the first conversation session (abstract: rephrasing the query, using a first large language model LLM, based on context from the chat history; col 3 l 10-21
col 15 l. 3-21: when the query is part of the conversation, method 800 can include activity 820 of rephrasing the query, using a first LLM, based on context from the chat history. In some embodiments, the first LLM can be a text-generation LLM, such as OpenAI's GPT-3, GPT 3.5, GPT4, and/or another suitable LLM. When the query is part of an existing chat, the query can be passed to the first LLM to rephrase the question to take into account the chat history. For example, assume the user previously asked for success enablers for people with anxiety disorder, and the chat recommended frequent breaks, emotional support animals, and apps for stress. If the user then asks, “Tell me more about the last one”, the first LLM can use this query and the chat history to rephrase the question to say, “Tell me about apps for stress” and pass use that updated query in activity 830, described below. In some cases, the chat history can be too lengthy to be included in the LLM prompt based on a limited input size of the LLM, in which case the chat history can first be summarized.;
col 15 l 22-33: prompt that can be used to the first LLM to generate the rephrased query);
determining that the rephrased query corresponds to a first domain {of a plurality of domains} (abstract: whether or not the query is related to accommodations or disabilities; fig 8 830; col 3 l 10-21; col 15 l. 34-41);
forwarding the rephrased query to a first service {of a plurality of services} that corresponds to the first domain (abstract: When the query is related to accommodations or disabilities, the method further can include determining one or more accommodations responsive to the query; col 3 l 10-21; col 15 l. 16-18; col 16 l. 41-46); and
based on obtaining a response to the rephrased query from the first service, updating the context of the first conversation session based on the response to the rephrased query (abstract: formulating a response to the query using a fourth LLM based on the one or more accommodations; col 3 l 10-21 conversation…chat history; col 15 l. 22-24; col 17 l. 58-61 – where chat history incorporates query and responses from user and system to allow the chat to use short-term memory in order to have a “conversation” with the user).
Thomas Does not specifically teach where Hamilton teaches a first domain of a plurality of domains; and a first service, of a plurality of services that corresponds to the first domain (0079: receive queries associated with multiple disparate subject matters and/or knowledge domains; 0113: query processing techniques…may be implemented for a plurality of different knowledge domains).
It would have been obvious to one of ordinary skill in the art before the effective filing date to incorporate Hamilton and multiple domains for an improved system, allowing for results to be provided for queries of multiple domains for a more comprehensive and efficient chat assistant of Thomas.
Thomas already teaches determining a response to a query of a particular domain. Similarly, Hamilton teaches using a machine learning framework to generate query results. One could thus look to Hamilton to further process query results in a first domain of a plurality of domains for a more comprehensive chat assistant, allowing for user to input queries on a plethora of topics and not face input restrictions.
Regarding claim 2 Thomas does not specifically teach where Hamilton teaches The method of claim 1, wherein updating the context of the first conversation session comprises prompting a second language model to extract a first subset of the information from the user query and to infer a second subset of the information from the user query.
([0003] In some embodiments, a computer-implemented method includes generating, by one or more processors and using a machine learning framework, one or more predictions for a natural language query, wherein the one or more predictions comprise (i) an intent prediction indicative of a likelihood of a target query intent and (ii) an event prediction indicative of a likelihood of a target event that is associated with the target query intent; generating, by the one or more processors, an intent classification for the natural language query based on the intent prediction and the event prediction
[0110] In some embodiments, the machine learning intent prediction model 318 has been fine-tuned over universal sentence encoding model, such as LLM universal sentence encoding model, to classify target query intent related query terms from other query terms seen by the predictive query system 302. In some examples, the machine learning intent prediction model 318 has been fine-tuned with training dataset 408 that includes a terminology corpus associated with the target query intent. In some examples, the training dataset includes supervised training data and/or unsupervised training data. The training dataset 408, may include a plurality of previous input query data objects. In some embodiments at least a subset of the plurality of previous input query data objects may be associated with a label (e.g., an intent class). For example, the training dataset may include labeled data, such as manually labeled data. In some examples, the training dataset 408 may include user-feedback data.).
It would have been obvious to one of ordinary skill in the art before the effective filing date to incorporate Hamilton for an improved system, allowing for the proper determination of the conversation context for improved response generation by the chat assistant.
Thomas already teaches receiving queries, chat history, and determining the context of a conversation, where Hamilton further teaches more specific query analysis with intent prediction to optimize query results (2-3). Thus, one could look to Hamilton to further incorporate the more specific query analysis to obtain information to better determine the context of the query for improved rephrasing and response generation.
Regarding claim 3 Thomas does not specifically teach where Hamilton teaches The method of claim 2, wherein prompting the second language model to extract the first subset of information and to infer the second subset of information from the user query comprises submitting a prompt to the second language model comprising the user query, a plurality of examples of user queries and corresponding information extracted from the user queries, and a task instruction to extract the first subset of information and infer the second subset of information from the user query based on the plurality of examples
([0003] In some embodiments, a computer-implemented method includes generating, by one or more processors and using a machine learning framework, one or more predictions for a natural language query, wherein the one or more predictions comprise (i) an intent prediction indicative of a likelihood of a target query intent and (ii) an event prediction indicative of a likelihood of a target event that is associated with the target query intent; generating, by the one or more processors, an intent classification for the natural language query based on the intent prediction and the event prediction
[0110] In some embodiments, the machine learning intent prediction model 318 has been fine-tuned over universal sentence encoding model, such as LLM universal sentence encoding model, to classify target query intent related query terms from other query terms seen by the predictive query system 302. In some examples, the machine learning intent prediction model 318 has been fine-tuned with training dataset 408 that includes a terminology corpus associated with the target query intent. In some examples, the training dataset includes supervised training data and/or unsupervised training data. The training dataset 408, may include a plurality of previous input query data objects. In some embodiments at least a subset of the plurality of previous input query data objects may be associated with a label (e.g., an intent class). For example, the training dataset may include labeled data, such as manually labeled data. In some examples, the training dataset 408 may include user-feedback data.).
It would have been obvious to one of ordinary skill in the art before the effective filing date to incorporate Hamilton for an improved system, allowing for the proper determination of the conversation context for improved response generation by the chat assistant.
Thomas already teaches determining the context of a conversation, and one could look to Hamilton to further incorporate the specific training and fine-tuning in obtaining most accurate query information to better determine the context of the query for improved rephrasing and response generation.
Regarding claim 4 Thomas teaches The method of claim 1, further comprising generating a response to the user query based on the response to the rephrased query and responding to the user query with the generated response (abstract: formulating a response to the query using a fourth LLM based on the one or more accommodations; fig 8; col 3 l 10-21).
Regarding claim 5 Thomas teaches The method of claim 1, wherein prompting the first language model comprises,
generating a prompt comprising the user query, the context of the first conversation session, and a task instruction to rephrase the user query based on the context of the first conversation session (col 15 l 15-16: the first LLM can use this query and the chat history to rephrase the question; l. 22-33: prompt); and
submitting the prompt to the first language model, wherein a response to the prompt obtained from the first language model comprises the rephrased query
(abstract: rephrasing the query, using a first large language model LLM, based on context from the chat history
Col 15 l. 3-21: when the query is part of the conversation, method 800 can include activity 820 of rephrasing the query, using a first LLM, based on context from the chat history. In some embodiments, the first LLM can be a text-generation LLM, such as OpenAI's GPT-3, GPT 3.5, GPT4, and/or another suitable LLM. When the query is part of an existing chat, the query can be passed to the first LLM to rephrase the question to take into account the chat history. For example, assume the user previously asked for success enablers for people with anxiety disorder, and the chat recommended frequent breaks, emotional support animals, and apps for stress. If the user then asks, “Tell me more about the last one”, the first LLM can use this query and the chat history to rephrase the question to say, “Tell me about apps for stress” and pass use that updated query in activity 830, described below. In some cases, the chat history can be too lengthy to be included in the LLM prompt based on a limited input size of the LLM, in which case the chat history can first be summarized;
Col 15 l. 22-33: Activity 820 can allow the chat to use short-term memory in order to have a “conversation” with the user, since LLMs are not inherently capable of having memory. In many embodiments, the first LLM is not fine-tuned for this task. An example of a prompt that can be used to the first LLM to generate the rephrased query is as follows (but this prompt is merely exemplary, and prompts can change from time to time based on the type of LLM being used): “Given the following conversation and a follow-up question, rephrase the question to be a standalone question.” [The chat history is included as the conversation, and the query is included as the follow-up question.]).
Regarding claim 7 Thomas teaches The method of claim 1, wherein determining that the user query corresponds to the first domain comprises determining intent of the user query and determining that the user query corresponds to the first domain based at least partly on the intent of the user query (col 15 l 35-36: determining, using a second LLM, whether or not the query is related to accommodations or disabilities.).
Regarding claim 8 Thomas teaches The method of claim 1, further comprising updating the context of the first conversation session to indicate the rephrased query (col 3 l 10-21 conversation…context…chat history; col 15 l. 22-24; col 17 l. 58-61 – where chat history incorporates query and responses from user and system to allow the chat to use short-term memory in order to have a “conversation” with the user).
Regarding claim 9 Thomas teaches The method of claim 1
wherein updating the context of the first conversation session based on the user query comprises updating a data structure with the information extracted and inferred from the user query (col 3 l 10-21 conversation…context…chat history; col 15 l. 22-24; col 17 l. 58-61),
wherein updating the context of the first conversation session based on the response to the rephrased query comprises updating the data structure with information that is at least one of extracted and inferred from the response to the rephrased query
(col 14 l 56- col 15 l 2: Referring to FIG. 8, method 800 can include an activity 810 of determining whether a query is part of a conversation having a chat history before the query. In some embodiments, the query can be a query received at employment system 310 (FIG. 3) from a user 350 (FIG. 3). For example, applicant 410 can use chat system 318 to input a query. In some cases, the query can be part of a larger conversation, such as an existing chat history. Activity 810 can involve determining whether the query is the first query submitted in a new conversation, the flow of method 800 can proceed to an activity 830, in which case the query, as submitted, can be used in activity 830, as described below. When the query is part of an existing chat, the flow of method 800 can proceed to an activity 820, as described below.
col 3 l 10-21 conversation…context…chat history; col 15 l. 22-24; col 17 l. 58-61).
Regarding claim 10 Thomas and Hamilton teach The method of claim 1, wherein each of the plurality of services interfaces with a corresponding one of a plurality of language models to generate responses to rephrased queries, wherein the first language model differs from the plurality of language models;
Where Thomas teaches
(col 15 l 35-36: determining, using a second LLM, whether or not the query is related to accommodations or disabilities.)
And Hamilton teaches plurality of services (79; 110, 113).
Rejected for similar rationale and reasoning as claim 1
Regarding claim 11 Thomas teaches One or more non-transitory machine-readable media having program code stored thereon, the program code comprising instructions (col 3 l. 22-26) to:
update a maintained state of a first conversation based on a first query comprising natural language that corresponds to the first conversation, wherein the instructions to update the maintained state of the first conversation comprise instructions to determine information from the first query and update the maintained state with the determined information (abstract; col 3 l 10-21 query is part of the conversation; context from the chat history; col 15 l. 3-21 “Tell me more about the last one”; Col 15 l. 22-33: Activity 820 can allow the chat to use short-term memory in order to have a “conversation” with the user, since LLMs are not inherently capable of having memory.; col 17 l. 58-61 chat history, kept in cache memory);
generate a rephrased query based on the first query and the maintained state of the first conversation, wherein the instructions to generate the rephrased query comprise instructions to prompt a first foundation model with the first query and the maintained state of the first conversation, wherein the rephrased query incorporates context of the first conversation identified from the maintained state of the first conversation (abstract: rephrasing the query, using a first large language model LLM, based on context from the chat history; col 3 l 10-21
col 15 l. 3-21: when the query is part of the conversation, method 800 can include activity 820 of rephrasing the query, using a first LLM, based on context from the chat history. In some embodiments, the first LLM can be a text-generation LLM, such as OpenAI's GPT-3, GPT 3.5, GPT4, and/or another suitable LLM. When the query is part of an existing chat, the query can be passed to the first LLM to rephrase the question to take into account the chat history. For example, assume the user previously asked for success enablers for people with anxiety disorder, and the chat recommended frequent breaks, emotional support animals, and apps for stress. If the user then asks, “Tell me more about the last one”, the first LLM can use this query and the chat history to rephrase the question to say, “Tell me about apps for stress” and pass use that updated query in activity 830, described below. In some cases, the chat history can be too lengthy to be included in the LLM prompt based on a limited input size of the LLM, in which case the chat history can first be summarized;
col 15 l. 22-33);
determine a destination of the rephrased query based on intent of the rephrased query, wherein the destination of the rephrased query corresponds to a first task resolver {of a plurality of task resolvers that corresponds to a plurality of task domains} (abstract: whether or not the query is related to accommodations or disabilities; fig 8 830; col 3 l 10-21; col 15 l. 34-41);
route the rephrased query to the determined destination to obtain a response to the rephrased query from the first task resolver (abstract: When the query is related to accommodations or disabilities, the method further can include determining one or more accommodations responsive to the query; col 3 l 10-21; col 15 l. 16-18); and
update the maintained state of the first conversation based on the response to the first query (abstract: formulating a response to the query using a fourth LLM based on the one or more accommodations; col 3 l 10-21 conversation…chat history
col 15 l. 22-24; col 17 l. 58-61 – where chat history incorporates query and responses from user and system to allow the chat to use short-term memory in order to have a “conversation” with the user).
Thomas however Does not specifically teach where Hamilton teaches:
a first task resolver of a plurality of task resolvers that corresponds to a plurality of task domains.
Rejected for similar rationale and reasoning as claim 1 where
(It would have been obvious to one of ordinary skill in the art before the effective filing date to incorporate Hamilton and multiple domains for an improved system, allowing for results to be provided for queries of multiple domains for a more comprehensive and efficient chat assistant of Thomas.)
Regarding claim 12 Thomas does not specifically teach where Hamilton teaches The non-transitory machine-readable media of claim 11
wherein the instructions to determine the information from the first query further comprise instructions to prompt a second foundation model to determine information from the first query,
wherein the instructions to prompt the second foundation model comprise instructions to submit a prompt to the second foundation model comprising the first query, a plurality of examples of queries comprising natural language and corresponding information determined from the queries, and a task instruction to determine the information from the first query based on the plurality of examples.
Rejected for similar rationale and reasoning as claims 2/3 above
Regarding claim 13 Thomas and Hamilton teach The non-transitory machine-readable media of claim 11, wherein the instructions to determine the destination of the rephrased query comprise instructions to determine that the rephrased query corresponds to a first task domain of the plurality of task domains based at least partly on the intent of the rephrased query, wherein the first task domain corresponds to the first task resolver (Thomas col 15 l 35-36: determining, using a second LLM, whether or not the query is related to accommodations or disabilities.).
Further rejected for similar rationale and reasoning as claim 11 (with Hamilton teaching the plurality of task domains)
Regarding claim 15 Thomas teaches An apparatus comprising:
a processor; and
a computer-readable medium having instructions stored thereon that are executable by the processor to cause the apparatus to (col 3 l 8-9 computer-implemented method; l 22-26: a system including one or more processors and one or more non-transitory computer-readable media storing computing instructions that, when executed on the one or more processors, cause the one or more processors to perform various operations),
during a first conversation that at least includes a plurality of user queries and a plurality of responses from at least one of a plurality of language models, for each user query of the plurality of user queries (abstract; col 3 l 10-21 query is part of the conversation; context from the chat history; col 15 l. 3-21 “Tell me more about the last one”; col 17 l. 58-61 chat history, kept in cache memory),
extract first information and infer second information from the user query (abstract; col 3 l 10-21 query is part of the conversation; context from the chat history; col 15 l. 3-21 “Tell me more about the last one”; col 17 l. 58-61 chat history, kept in cache memory);
update a context for the first conversation with the first and second information to generate an updated context for the first conversation (abstract; col 3 l 10-21 query is part of the conversation; context from the chat history; col 15 l. 3-21 “Tell me more about the last one”; col 17 l. 58-61 chat history, kept in cache memory);
rephrase the user query based on the updated context to generate a rephrased user query (abstract: rephrasing the query, using a first large language model LLM, based on context from the chat history; col 3 l 10-21
col 15 l. 3-21: when the query is part of the conversation, method 800 can include activity 820 of rephrasing the query, using a first LLM, based on context from the chat history. In some embodiments, the first LLM can be a text-generation LLM, such as OpenAI's GPT-3, GPT 3.5, GPT4, and/or another suitable LLM. When the query is part of an existing chat, the query can be passed to the first LLM to rephrase the question to take into account the chat history. For example, assume the user previously asked for success enablers for people with anxiety disorder, and the chat recommended frequent breaks, emotional support animals, and apps for stress. If the user then asks, “Tell me more about the last one”, the first LLM can use this query and the chat history to rephrase the question to say, “Tell me about apps for stress” and pass use that updated query in activity 830, described below. In some cases, the chat history can be too lengthy to be included in the LLM prompt based on a limited input size of the LLM, in which case the chat history can first be summarized.;
col 15 l 22-33: prompt that can be used to the first LLM to generate the rephrased query);
determine a task domain based on the rephrased user query (abstract: whether or not the query is related to accommodations or disabilities; fig 8 830; col 3 l 10-21; col 15 l. 34-41);
identify a first service {of a plurality of services} to handle the rephrased user query based on the task domain, wherein each {of the plurality of} service{s} corresponds to one of the plurality of language models (abstract: When the query is related to accommodations or disabilities, the method further can include determining one or more accommodations responsive to the query; col 3 l 10-21; col 15 l. 16-18, l 34-37; col 16 l. 41-46); and
communicate the rephrased user query to the first service to obtain in response a corresponding one of the plurality of responses (abstract: formulating a response to the query using a fourth LLM based on the one or more accommodations); and
for each response of the plurality of responses, update the context for the first conversation based on the response query (abstract: formulating a response to the query using a fourth LLM based on the one or more accommodations; col 3 l 10-21 conversation…chat history; col 15 l. 22-24; col 17 l. 58-61 – where chat history incorporates query and responses from user and system to allow the chat to use short-term memory in order to have a “conversation” with the user).
However, Thomas does not specifically teach where Hamilton teaches a first service of a plurality of services
Rejected for similar rationale and reasoning as claim 1 where
(It would have been obvious to one of ordinary skill in the art before the effective filing date to incorporate Hamilton and multiple domains for an improved system, allowing for results to be provided for queries of multiple domains for a more comprehensive and efficient chat assistant of Thomas.)
Claims 16-17 recite limitations similar to claims 2-3 and are Rejected for similar rationale and reasoning as claims 2/3 above
Regarding claim 18 Thomas teaches The apparatus of claim 15, wherein the instructions executable by the processor to cause the apparatus to determine the task domain based on the rephrased user query comprise instructions to determine the task domain based at least partly on a determined intent of the rephrased user query (col 15 l 35-36: determining, using a second LLM, whether or not the query is related to accommodations or disabilities.).
Regarding claim 20 Thomas teaches The apparatus of claim 15,
wherein the instructions executable by the processor to cause the apparatus to update the context for the first conversation comprise instructions executable by the processor to cause the apparatus to update a data structure with the first and second information (col 3 l 10-21 conversation…context…chat history; col 15 l. 22-24; col 17 l. 58-61),
wherein the instructions executable by the processor to cause the apparatus to update the context for the first conversation based on the response comprise instructions executable by the processor to cause the apparatus to update the data structure to indicate the response
(col 14 l 56- col 15 l 2: Referring to FIG. 8, method 800 can include an activity 810 of determining whether a query is part of a conversation having a chat history before the query. In some embodiments, the query can be a query received at employment system 310 (FIG. 3) from a user 350 (FIG. 3). For example, applicant 410 can use chat system 318 to input a query. In some cases, the query can be part of a larger conversation, such as an existing chat history. Activity 810 can involve determining whether the query is the first query submitted in a new conversation, the flow of method 800 can proceed to an activity 830, in which case the query, as submitted, can be used in activity 830, as described below. When the query is part of an existing chat, the flow of method 800 can proceed to an activity 820, as described below.
col 3 l 10-21 conversation…context…chat history; col 15 l. 22-24; col 17 l. 58-61).
7. Claims 6, 14, and 19 are rejected under 35 U.S.C. 103 as being unpatentable over Thomas in view of Hamilton in further view of Lee et al (2025/0298817).
Regarding claim 6 Thomas does not specifically teach where Lee teaches The method of claim 5, wherein the prompt further comprises a plurality of examples of user queries and rephrased queries based on corresponding contexts of conversation sessions
([0075] The meta prompt optimization engine 440 is configured to provide an optimized meta prompt for the task prompt enhancement engine 430. The meta prompt optimization engine 440 can include a meta prompt rephrasing module 450, a task prompt enhancing module 460, and an evaluation module 470. The meta prompt rephrasing module 450 is configured to generate multiple variant meta prompts for an initial meta prompt. The meta prompt rephrasing module 450 can implement or use a trained LLM from a model store 420 on the communication platform 310 or from a remote server 380. An initial meta prompt can be the input of the trained LLM. The initial meta prompt can be predefined, for example by an operator of the communication platform 310, as a default prompt for enhancing user provided task prompts. The trained LLM can rephrase or paraphrase an input, for example the initial meta prompt, based on an instruction (which is a prompt to the trained LLM). For example, the instruction for the trained LLM is to rephrase the initial meta prompt to generate five variants. The number of the variants can be adjusted by an operator of the communication platform 310. Certain parameters in the trained LLM for meta prompt rephrasing can be adjusted to diversify the variants. For example, a temperature parameter of the trained LLM can be set in a range between 0 and 1. The higher the temperature parameter is, the less deterministic the generated result can become, thus the variant meta prompts generated can be more diversified and different from the initial meta prompt. In some examples, the trained LLM for meta prompt rephrasing uses evaluation data associated with previously generated variant meta prompts from the evaluation module 470, which will be described below, as feedback input, for example for refinement and fine-tuning.).
It would have been obvious to one of ordinary skill in the art before the effective filing date to incorporate Lee for an improved system, allowing for the proper rephrasing of the query.
Thomas already teaches rephrasing of the query, and one could look to further incorporate Lee for refinement and fine-tuning (0075) of the LLM designated for rephrasing, and ultimately improving rephrasing and response generation for the chat assistant of Thomas.
Regarding claim 14 Thomas, Hamilton, and Lee teach The non-transitory machine-readable media of claim 11, wherein the instructions to prompt the first foundation model comprise instructions to,
generate a prompt comprising the first query, the maintained state of the first conversation, a plurality of examples of queries comprising natural language and rephrased versions of the queries based on corresponding conversation states, and a task instruction to rephrase the first query based on context provided by the maintained state of the first conversation and the plurality of examples; and
submit the prompt to the first foundation model, wherein a response to the prompt obtained from the first foundation model comprises the rephrased query.
Recites limitations similar to claims 5-6 and is Rejected for similar rationale and reasoning
Regarding claim 19 Thomas, Hamilton, and Lee teach The apparatus of claim 15, wherein the instructions executable by the processor to cause the apparatus to rephrase the user query comprise instructions executable by the processor to cause the apparatus to submit a prompt to a language model comprising the user query, the updated context, one or more examples of user queries and rephrased versions of the user queries based on corresponding conversation contexts, and a task instruction to rephrase the user query based on the updated context and the one or more examples.
Recites limitations similar to claims 5-6 and is Rejected for similar rationale and reasoning
Conclusion
8. The prior art made of record and not relied upon is considered pertinent to applicant's disclosure: See PTO-892.
Cheng et al (2025/0372090)
Abstract: Dialogue state tracking for voice assistants involves correctly tracking intent and entities of a task that a user is performing. A dialogue state, having a tracked intent and one or more tracked entities, can then be used to perform the task. Building a dialogue state tracking system within a voice assistant is not trivial. In some embodiments, a dialogue state tracking system involving one or more large language models can be implemented downstream of a natural language understanding system to produce the tracked intent and the one or more tracked entities. In some embodiments, a dialogue state tracking system involving one or more large language models can be implemented upstream of a natural language understanding system to produce rephrased natural language text, which is in turn processed by the natural language understanding system to produce the tracked intent and the one or more tracked entities.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to SHAUN A ROBERTS whose telephone number is (571)270-7541. The examiner can normally be reached Monday-Friday 9-5 EST.
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, Andrew Flanders can be reached on 571-272-7516. 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.
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 or access to the automated information system, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/SHAUN ROBERTS/Primary Examiner, Art Unit 2655
Prosecution Insights