DETAILED ACTION
Receipt of Applicant’s Amendment, filed July 13, 2026 is acknowledged.
Claims 1-3, 5-7, 12-14, and 19 were amended.
Claims 8-11, 16, 18, and 20 are withdrawn
Claims 1-3, 5-7, 12-14, 17, and 19 are pending in this office action.
Specification
Applicant is reminded of the proper language and format for an abstract of the disclosure.
The abstract should be in narrative form and generally limited to a single paragraph on a separate sheet within the range of 50 to 150 words in length. The abstract should describe the disclosure sufficiently to assist readers in deciding whether there is a need for consulting the full patent text for details.
The language should be clear and concise and should not repeat information given in the title. It should avoid using phrases which can be implied, such as, “The disclosure concerns,” “The disclosure defined by this invention,” “The disclosure describes,” etc. In addition, the form and legal phraseology often used in patent claims, such as “means” and “said,” should be avoided.
The instant abstract a single sentence not a paragraph and is written in claim form and uses legal phraseology. It is suggested that the abstract be amended to be written in a narrative format.
Claim Rejections - 35 USC § 102
The following is a quotation of the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(1) the claimed invention was patented, described in a printed publication, or in public use, on sale, or otherwise available to the public before the effective filing date of the claimed invention.
Claims 1, 7, 12, 17, and 19 are rejected under 35 U.S.C. 102(a)(1) as being anticipated by RodriguezGarcia [2025/0307564].
With regard to claim 1 RodriguezGarcia teaches A method for intent recognition based on a large language model (LLM) (RodriguezGarcia, ¶30 “In some implementations, both neural networks 106 and 124 may be large language models (LLMs), which have undergone extensive training on vast and diverse text corpora encompassing a wide range of domains and context.”), comprising:
obtaining a query statement as obtaining an utterance (RodriguezGarcia, ¶3 “method of intent discovery includes obtaining an utterance and a corresponding label representative of an intent of the utterance”), a preset intent as the corresponding label of known intent (id; ¶24 “a list of known intents pertinent to the domain. These known intents may be retrieved from a database. Additionally, each prompt may include selected examples ( e.g., utterances) from a training dataset to provide context to the prompt”), and descriptive information as the task description (RodriguezGarcia, ¶25 “a prompt comprising a task description and an input label pair”; ¶52 “the instructions 302 may be a task description that instructs a prompt generator (e.g., the prompt generator 106 described in FIG. 1) to generate a prompt for use by another LLM (e.g., the intent predictor 124 described in FIG. 1).”) of the preset intent as the known intents (RodriguezGarcia, ¶41);
obtaining, from the present intent as the known intents (RodriguezGarcia, ¶41), a first candidate intent as the discovered intent (RodriguezGarcia, ¶53 “The prompt 400 may include instructions 402 and 406 instructing an LLM (e.g., the intent predictor 124) to discover intents for utterances using samples 404 as a guide”; Figure 1, 126; Figure 2, 220) corresponding to the query statement as for the utterance (RodriguezGarcia, ¶39 “here each utterance in both the few-shot pool 108 and the test batch”) by matching (RodriguezGarcia, ¶39 “based on a similarity measure (e.g., cosine distance) between the one or more examples and their respective test-batch utterance.”; ¶57 “If the utterance matches a known intent, assign it to that intent.”; ¶84 “which may use a pre-trained neural network model or another mechanism to determine whether an intent matching the utterance exists in a database 810 that stores known intents. If a match is found, the intent retriever 808 responds to the IVR module 806 with the matching intent.”; ¶94 “In some implementations, utterances corresponding to the list of known intents are semantically similar to utterances in the list of test examples. In some implementations, an utterance in the few-shot example is semantically similar to an utterance in at least one of the list of test examples.”) the query statement as the utterance in the test examples (Id) with the preset intent as the list of known intents (Id) and the preset descriptive information (RodriguezGarcia, ¶52 “the instructions 302 may be a task description that instructs a prompt generator (e.g., the prompt generator 106 described in FIG. 1) to generate a prompt for use by another LLM (e.g., the intent predictor 124 described in FIG. 1).”) of the preset intent as the known intents (RodriguezGarcia, ¶41);
generating first prompt information as prompt 208 (RodriguezGarcia, ¶50 “the one or more software components augment the prompt 208 generated by the process 200A with samples from few-shot samples 212 and intents from known intents 214 to create an augmented prompt 216.”) based on the query statement as the utterance in the test examples (RodriguezGarcia, ¶3 “method of intent discovery includes obtaining an utterance and a corresponding label representative of an intent of the utterance”), the first candidate intent as the discovered intent (RodriguezGarcia, ¶53 “The prompt 400 may include instructions 402 and 406 instructing an LLM (e.g., the intent predictor 124) to discover intents for utterances using samples 404 as a guide”; Figure 1, 126; Figure 2, 220), and descriptive information as the task description (RodriguezGarcia, ¶25; ¶52; Please note this claim limitation has been construed to mean --first descriptive information--). of the first candidate intent as the discovered intent (RodriguezGarcia, ¶53; Figure 1, 126; Figure 2, 220);
obtaining a first intent as the discovered intent (RodriguezGarcia, ¶53 “The prompt 400 may include instructions 402 and 406 instructing an LLM (e.g., the intent predictor 124) to discover intents for utterances using samples 404 as a guide”) corresponding to the query statement as for the utterance (RodriguezGarcia, ¶39 “here each utterance in both the few-shot pool 108 and the test batch”) from the first candidate intent as the discovered intent (RodriguezGarcia, ¶53 “The prompt 400 may include instructions 402 and 406 instructing an LLM (e.g., the intent predictor 124) to discover intents for utterances using samples 404 as a guide”) by inputting the first prompt information as prompt 208 (RodriguezGarcia, ¶50) into the LLM (RodriguezGarcia, ¶30; Figure 2, 206; 218);
generating second prompt information as augmented prompt 216 (RodriguezGarcia, Figure 2B, 216) based on the first intent as a new discovered intent, discovered during iteration (RodriguezGarcia, ¶53) and the first prompt information as prompt 208 (RodriguezGarcia, ¶50 “the one or more software components augment the prompt 208 generated by the process 200A with samples from few-shot samples 212 and intents from known intents 214 to create an augmented prompt 216.”);
determining a second intent as a second discovered intent (RodriguezGarcia, Figure 2B, 220) corresponding to the query statement as the utterance in the test examples (RodriguezGarcia, ¶3 “method of intent discovery includes obtaining an utterance and a corresponding label representative of an intent of the utterance”) from another first candidate intent as the list of known intents selected (RodriguezGarcia, ¶70 “In operation 606, the processing logic modifies the prompt to include one or more of: a list of known intents, a few-shot example, or a list of test examples, as shown in FIG. 2B. The list of known intents includes one or more intents from a training dataset and one or more intents discovered by the second large language model in one or more previous iterations. The few-shot example may be selected from a few-shot pool that was prepopulated with utterance-intent pairs extracted from the training dataset.”) except for the first intent as the intents not selected for the few-shot pool (Id) by inputting the second prompt information as augmented prompt 216 (RodriguezGarcia, Figure 2B, 216) into the LLM (RodriguezGarcia, ¶30; Figure 2, 206; 218), a first processing operation as update with Novel Intents 222 for the first intent in the first iteration (RodriguezGarcia, Figure 2B; ¶6; ¶46; Please see the 112b above, this claim limitation has been construed to man –a second processing operation--) corresponding to the first intent as the first discovered intent (RodriguezGarcia, ¶53 “The prompt 400 may include instructions 402 and 406 instructing an LLM (e.g., the intent predictor 124) to discover intents for utterances using samples 404 as a guide”; Figure 1, 126; Figure 2, 220) in the query statement as for the utterance (RodriguezGarcia, ¶39), and wherein a second processing operation as update with Novel Intents 222 for the second intent in a second iteration (RodriguezGarcia, Figure 2B; ¶6 “the list of known intents includes at one intent from a training dataset and at least one intent discovered by the second large language model in a previous iteration.”; ¶46 “These expanded known intents may be used as contextual information by the intent predictor 124 to discover (generate) intents on utterances in subsequent iterations.”) corresponds to the second intent as a second discovered intent (RodriguezGarcia, Figure 2B, 220) in the query statement as the utterance in the test examples (RodriguezGarcia, ¶3 “method of intent discovery includes obtaining an utterance and a corresponding label representative of an intent of the utterance”) is executed after a the first processing operation as update with Novel Intents 222 for the first intent in the first iteration (RodriguezGarcia, Figure 2B; ¶6; ¶46
generating new second prompt information as updating the known intents with the newly discovered intent (RodriguezGarcia, ¶46 “the intent discovery system 102 may update the known intents 112 with the newly discovered intents, which are intent predictions 126, to expand the known intents.”) based on the second intent as a second discovered intent (RodriguezGarcia, Figure 2B, 220) and the second prompt information as augmented prompt 216 (RodriguezGarcia, Figure 2B, 216), and returning to perform an operation as iterate (RodriguezGarcia, ¶70) of obtaining the second intent as discovering intents (RodriguezGarcia, Figure 2B, 220) until the LLM (RodriguezGarcia, ¶30; Figure 2, 206; 218) outputs termination indication information as when the LLM has completed the task, and has updated the list of known intents with a new intent (RodriguezGarcia, Figure 6, 612; Figure 7 710); and
determining the first intent and the second intent as the first target intent as a new discovered intent, discovered during iteration which are stored (RodriguezGarcia, ¶53 “The prompt 400 may include instructions 402 and 406 instructing an LLM (e.g., the intent predictor 124) to discover intents for utterances using samples 404 as a guide”; ¶24 “Identified intents that are not in the list of known intents are used to update the database. Through multiple iterations with varying test data, the database progressively expands its repository of known intents related to each domain.”).
With regard to claim 7 RodriguezGarcia further teaches in response to the preset descriptive information as the task description (RodriguezGarcia, ¶25 “a prompt comprising a task description and an input label pair”; ¶52 “the instructions 302 may be a task description that instructs a prompt generator (e.g., the prompt generator 106 described in FIG. 1) to generate a prompt for use by another LLM (e.g., the intent predictor 124 described in FIG. 1).”) comprising a parameter type as instructions regarding classifications (RodriguezGarcia, ¶61) corresponding to the preset intent as the corresponding label of known intent (RodriguezGarcia, ¶3; ¶24 and first definition information as the predefined contents and format template (RodriguezGarcia, ¶68 “The first large language model generates the prompt based on a pre-determined template that defines contents and format of the prompt.”) corresponding to the parameter type as the template containing the classification instructions (RodriguezGarcia, ¶61), generating third prompt information as a second augmented prompt 216 during iteration (RodriguezGarcia, Figure 2B, 216; ¶70) based on the query statement as for the utterance (RodriguezGarcia, ¶39), the first candidate intent as the first discovered intent (RodriguezGarcia, ¶53), and the descriptive information as the first task description (RodriguezGarcia, ¶25; ¶52) corresponding to the first candidate intent as the first discovered intent (RodriguezGarcia, ¶53), wherein the third prompt information as a second augmented prompt 216 during iteration (RodriguezGarcia, Figure 2B, 216; ¶70) is configured to prompt the LLM (RodriguezGarcia, ¶30; Figure 2, 206; 218) to determine the first target intent as a new discovered intent, discovered during iteration (RodriguezGarcia, ¶53; ¶24) corresponding to the query statement as the utterance in the test examples (RodriguezGarcia, ¶3) from the first candidate intent as the discovered intent which becomes a known intent (RodriguezGarcia, ¶53; ¶24) and extract a target parameter as instructions regarding classifications (RodriguezGarcia, ¶61) corresponding to the first target intent as a new discovered intent, discovered during iteration (RodriguezGarcia, ¶53; ¶24) from the query statement as the utterance in the test examples (RodriguezGarcia, ¶3);
obtaining the first target intent as a new discovered intent, discovered during iteration (RodriguezGarcia, ¶53; ¶24)corresponding to the query statement as the utterance in the test examples (RodriguezGarcia, ¶3) and the target parameter as instructions regarding classifications (RodriguezGarcia, ¶61) associated with the first target intent as a new discovered intent, discovered during iteration (RodriguezGarcia, ¶53; ¶24) in the query statement as the utterance in the test examples (RodriguezGarcia, ¶3) by inputting the third prompt information as a second augmented prompt 216 during iteration (RodriguezGarcia, Figure 2B, 216; ¶70) into the LLM(RodriguezGarcia, ¶30; Figure 2, 206; 218); and
determining a response statement (RodriguezGarcia, ¶64 “a response format instruction 518 that instructs to the LLM to respond in a particular format. For example, the response format instruction 518 may state: "RESPONSE FORMAT: ID: <i>, Utterance: <content>, Intent: <intent> Use the same ID in the test example."”) corresponding to the query statement as the utterance in the test examples (RodriguezGarcia, ¶3) based on the first target intent as a new discovered intent, discovered during iteration (RodriguezGarcia, ¶53; ¶24) and the target parameter as instructions regarding classifications (RodriguezGarcia, ¶61).
With regard to claim 12 RodriguezGarcia teaches An electronic device, comprising:
at least one processor as a processor (RodriguezGarcia, ¶5” In some implementations, a system for intent discovery comprises one or more processors and memory including computer-executable instructions. The one or more processors, when executing computer-executable instructions, cause the system to perform operations that comprises”); and
a memory as a memory (Id) communicatively coupled to the at least one processor (Id),
wherein the processor (Id) is configured to:
obtain a query statement as obtaining an utterance (RodriguezGarcia, ¶3 “method of intent discovery includes obtaining an utterance and a corresponding label representative of an intent of the utterance”), a preset intent as the corresponding label of known intent (id; ¶24 “a list of known intents pertinent to the domain. These known intents may be retrieved from a database. Additionally, each prompt may include selected examples ( e.g., utterances) from a training dataset to provide context to the prompt”), and descriptive information as the task description (RodriguezGarcia, ¶25 “a prompt comprising a task description and an input label pair”; ¶52 “the instructions 302 may be a task description that instructs a prompt generator (e.g., the prompt generator 106 described in FIG. 1) to generate a prompt for use by another LLM (e.g., the intent predictor 124 described in FIG. 1).”) of the preset intent as the known intents (RodriguezGarcia, ¶41);
obtain, from the present intent as the known intents (RodriguezGarcia, ¶41), a first candidate intent as the discovered intent (RodriguezGarcia, ¶53 “The prompt 400 may include instructions 402 and 406 instructing an LLM (e.g., the intent predictor 124) to discover intents for utterances using samples 404 as a guide”; Figure 1, 126; Figure 2, 220) corresponding to the query statement as for the utterance (RodriguezGarcia, ¶39 “here each utterance in both the few-shot pool 108 and the test batch”) by matching (RodriguezGarcia, ¶39 “based on a similarity measure (e.g., cosine distance) between the one or more examples and their respective test-batch utterance.”; ¶57 “If the utterance matches a known intent, assign it to that intent.”; ¶84 “which may use a pre-trained neural network model or another mechanism to determine whether an intent matching the utterance exists in a database 810 that stores known intents. If a match is found, the intent retriever 808 responds to the IVR module 806 with the matching intent.”; ¶94 “In some implementations, utterances corresponding to the list of known intents are semantically similar to utterances in the list of test examples. In some implementations, an utterance in the few-shot example is semantically similar to an utterance in at least one of the list of test examples.”) the query statement as the utterance in the test examples (Id) with the preset intent as the list of known intents (Id) and the preset descriptive information (RodriguezGarcia, ¶52 “the instructions 302 may be a task description that instructs a prompt generator (e.g., the prompt generator 106 described in FIG. 1) to generate a prompt for use by another LLM (e.g., the intent predictor 124 described in FIG. 1).”) of the preset intent as the known intents (RodriguezGarcia, ¶41);
generate first prompt information as prompt 208 (RodriguezGarcia, ¶50 “the one or more software components augment the prompt 208 generated by the process 200A with samples from few-shot samples 212 and intents from known intents 214 to create an augmented prompt 216.”) based on the query statement as the utterance in the test examples (RodriguezGarcia, ¶3 “method of intent discovery includes obtaining an utterance and a corresponding label representative of an intent of the utterance”), the first candidate intent as the discovered intent (RodriguezGarcia, ¶53 “The prompt 400 may include instructions 402 and 406 instructing an LLM (e.g., the intent predictor 124) to discover intents for utterances using samples 404 as a guide”; Figure 1, 126; Figure 2, 220), and descriptive information as the task description (RodriguezGarcia, ¶25; ¶52; Please note this claim limitation has been construed to mean --first descriptive information--). of the first candidate intent as the discovered intent (RodriguezGarcia, ¶53; Figure 1, 126; Figure 2, 220);
obtain a first intent as the discovered intent (RodriguezGarcia, ¶53 “The prompt 400 may include instructions 402 and 406 instructing an LLM (e.g., the intent predictor 124) to discover intents for utterances using samples 404 as a guide”) corresponding to the query statement as for the utterance (RodriguezGarcia, ¶39 “here each utterance in both the few-shot pool 108 and the test batch”) from the first candidate intent as the discovered intent (RodriguezGarcia, ¶53 “The prompt 400 may include instructions 402 and 406 instructing an LLM (e.g., the intent predictor 124) to discover intents for utterances using samples 404 as a guide”) by inputting the first prompt information as prompt 208 (RodriguezGarcia, ¶50) into the LLM (RodriguezGarcia, ¶30; Figure 2, 206; 218);
generate second prompt information as augmented prompt 216 (RodriguezGarcia, Figure 2B, 216) based on the first intent as a new discovered intent, discovered during iteration (RodriguezGarcia, ¶53) and the first prompt information as prompt 208 (RodriguezGarcia, ¶50 “the one or more software components augment the prompt 208 generated by the process 200A with samples from few-shot samples 212 and intents from known intents 214 to create an augmented prompt 216.”);
determine a second intent as a second discovered intent (RodriguezGarcia, Figure 2B, 220) corresponding to the query statement as the utterance in the test examples (RodriguezGarcia, ¶3 “method of intent discovery includes obtaining an utterance and a corresponding label representative of an intent of the utterance”) from another first candidate intent as the list of known intents selected (RodriguezGarcia, ¶70 “In operation 606, the processing logic modifies the prompt to include one or more of: a list of known intents, a few-shot example, or a list of test examples, as shown in FIG. 2B. The list of known intents includes one or more intents from a training dataset and one or more intents discovered by the second large language model in one or more previous iterations. The few-shot example may be selected from a few-shot pool that was prepopulated with utterance-intent pairs extracted from the training dataset.”) except for the first intent as the intents not selected for the few-shot pool (Id) by inputting the second prompt information as augmented prompt 216 (RodriguezGarcia, Figure 2B, 216) into the LLM (RodriguezGarcia, ¶30; Figure 2, 206; 218), wherein a first processing operation as update with Novel Intents 222 for the first intent in the first iteration (RodriguezGarcia, Figure 2B; ¶6; ¶46; Please see the 112b above, this claim limitation has been construed to man –a second processing operation--) corresponding to the first intent as the first discovered intent (RodriguezGarcia, ¶53 “The prompt 400 may include instructions 402 and 406 instructing an LLM (e.g., the intent predictor 124) to discover intents for utterances using samples 404 as a guide”; Figure 1, 126; Figure 2, 220) in the query statement as for the utterance (RodriguezGarcia, ¶39), and a second processing operation as update with Novel Intents 222 for the second intent in a second iteration (RodriguezGarcia, Figure 2B; ¶6 “the list of known intents includes at one intent from a training dataset and at least one intent discovered by the second large language model in a previous iteration.”; ¶46 “These expanded known intents may be used as contextual information by the intent predictor 124 to discover (generate) intents on utterances in subsequent iterations.”) corresponds to the second intent as a second discovered intent (RodriguezGarcia, Figure 2B, 220) in the query statement as the utterance in the test examples (RodriguezGarcia, ¶3 “method of intent discovery includes obtaining an utterance and a corresponding label representative of an intent of the utterance”) is executed after a the first processing operation as update with Novel Intents 222 for the first intent in the first iteration (RodriguezGarcia, Figure 2B; ¶6; ¶46
generate new second prompt information as updating the known intents with the newly discovered intent (RodriguezGarcia, ¶46 “the intent discovery system 102 may update the known intents 112 with the newly discovered intents, which are intent predictions 126, to expand the known intents.”) based on the second intent as a second discovered intent (RodriguezGarcia, Figure 2B, 220) and the second prompt information as augmented prompt 216 (RodriguezGarcia, Figure 2B, 216), and returning to perform an operation as iterate (RodriguezGarcia, ¶70) of obtaining the second intent as discovering intents (RodriguezGarcia, Figure 2B, 220) until the LLM (RodriguezGarcia, ¶30; Figure 2, 206; 218) outputs termination indication information as when the LLM has completed the task, and has updated the list of known intents with a new intent (RodriguezGarcia, Figure 6, 612; Figure 7 710); and
determine the first intent and the second intent as the first target intent as a new discovered intent, discovered during iteration which are stored (RodriguezGarcia, ¶53 “The prompt 400 may include instructions 402 and 406 instructing an LLM (e.g., the intent predictor 124) to discover intents for utterances using samples 404 as a guide”; ¶24 “Identified intents that are not in the list of known intents are used to update the database. Through multiple iterations with varying test data, the database progressively expands its repository of known intents related to each domain.”).
With regard to claim 17 RodriguezGarcia further teaches A non-transitory computer readable storage medium storing computer instructions, wherein the computer instructions (RodriguezGarcia, ¶5” In some implementations, a system for intent discovery comprises one or more processors and memory including computer-executable instructions. The one or more processors, when executing computer-executable instructions, cause the system to perform operations that comprises”) are configured to cause a computer to implement the method of claim 1 (Please see the mapping for claim 1).
With regard to claim 19 RodriguezGarcia further teaches A computer program product comprising a processor and computer instructions, wherein when the computer instructions are executed by the processor (RodriguezGarcia, ¶5” In some implementations, a system for intent discovery comprises one or more processors and memory including computer-executable instructions. The one or more processors, when executing computer-executable instructions, cause the system to perform operations that comprises”), steps of the method of claim 1 are implemented (Please see the mapping for claim 1).
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 2, 3, 5, 6, 13 and 14 are rejected under 35 U.S.C. 103 as being unpatentable over RodriguezGarcia in view of Kharbanda [20240362279].
With regard to claims 2 and 13 RodriguezGarcia further teaches wherein generating first prompt information as prompt 208 (RodriguezGarcia, ¶50 “the one or more software components augment the prompt 208 generated by the process 200A with samples from few-shot samples 212 and intents from known intents 214 to create an augmented prompt 216.”) based on the query statement as the utterance in the test examples (RodriguezGarcia, ¶3 “method of intent discovery includes obtaining an utterance and a corresponding label representative of an intent of the utterance”), the first candidate intent as the discovered intent (RodriguezGarcia, ¶53 “The prompt 400 may include instructions 402 and 406 instructing an LLM (e.g., the intent predictor 124) to discover intents for utterances using samples 404 as a guide”; Figure 1, 126; Figure 2, 220), and first descriptive information as the task description (RodriguezGarcia, ¶25; ¶52; Please note this claim limitation has been construed to mean --first descriptive information--). of the first candidate intent as the discovered intent (RodriguezGarcia, ¶53; Figure 1, 126; Figure 2, 220) comprises:
obtaining a user as the user who makes the utterance (RodriguezGarcia, ¶24 “In some implementations, the test data may be real-world user utterances whose intents needs to be identified.”) [[ as the utterance being made (RodriguezGarcia, ¶24 “In some implementations, the test data may be real-world user utterances whose intents needs to be identified.”);
determining historical [[ as the known intents (RodriguezGarcia, ¶8 “In some implementations, the known intents are a subset of a plurality of known intents stored in a training dataset, where the training dataset includes a plurality of utterance-intent pairs of a particular domain.”) corresponding to the query statement as utterance (RodriguezGarcia, ¶8 “In some implementations, the known intents are a subset of a plurality of known intents stored in a training dataset, where the training dataset includes a plurality of utterance-intent pairs of a particular domain.”) based on the user [[ as the user who makes the utterance (RodriguezGarcia, ¶24); and
generating the first prompt information as prompt 208 (RodriguezGarcia, ¶50 “the one or more software components augment the prompt 208 generated by the process 200A with samples from few-shot samples 212 and intents from known intents 214 to create an augmented prompt 216.”) based on the query statement as the utterance in the test examples (RodriguezGarcia, ¶3 “method of intent discovery includes obtaining an utterance and a corresponding label representative of an intent of the utterance”), the first candidate intent as the discovered intent (RodriguezGarcia, ¶53 “The prompt 400 may include instructions 402 and 406 instructing an LLM (e.g., the intent predictor 124) to discover intents for utterances using samples 404 as a guide”; Figure 1, 126; Figure 2, 220), the first descriptive information of the first candidate intent as the discovered intent (RodriguezGarcia, ¶53; Figure 1, 126; Figure 2, 220), and the historical [[ as the known intents (RodriguezGarcia, ¶8).
RodriguezGarcia does not explicitly teach obtain a user identifier of the query statement; determining historical interaction information corresponding to the query statement based on the user identifier; and generating the first prompt information based on …and the historical interaction information.
Obtain a user identifier of the query statement;
Kharbanda teaches determining historical interaction information as the context which includes previous interaction data (¶130 “The context determination block 62 may identify and/or process metadata, user profile data (e.g., preferences, user search history, user browsing history, user purchase history, and/or user input data), previous interaction data”) corresponding to the query statement as the search is based on the context (¶39 “the system can provide more accurate search results by enhancing the query with additional signals that provide helpful context for the search.”) based on the user identifier as the user profile associated with a particular user (¶130 “The context determination block 62 may identify and/or process metadata, user profile data … associated with the user. The context can be associated… associated with the user and/or the retrieved or obtained data.”); and
generating the first prompt information as query refinement, which may include prompt generation (¶141 “For example, a search prompt, a purchase prompt, a generate prompt, a reservation prompt, a call prompt, a redirect prompt, and/or one or more other prompts may be determined to be associated with the output(s) of the sensor processing system 60”; ¶43 “In the context of search engines, query refinement can involve determining a user's intent by evaluating user input (e.g., image input 102, audio input 104) and refining the search query based on the user's intent.”) based on …and the historical interaction information as query refinement based on the context, e.g. the previous interactions (43 “In the context of search engines, query refinement can involve determining a user's intent by evaluating user input (e.g., image input 102, audio input 104) and refining the search query based on the user's intent.”; ¶130 “The context determination block 62 may identify and/or process metadata, user profile data (e.g., preferences, user search history, user browsing history, user purchase history, and/or user input data), previous interaction data”).
It would have been obvious to one of ordinary skill to which said subject matter pertains at the time the invention was filed to have implemented the search system taught by Kharbana to refine the query based on a user’s intent that is determined using the system taught by RodriguezGarcia as it yields the predictable results of enabling the system to use an expanded repository of known intents related to each domain (RodriguezGarcia, ¶24 “Through multiple iterations with varying test data, the database progressively expands its repository of known intents related to each domain. In some implementations, the test data may be real-world user utterances whose intents needs to be identified”). The search system taught by Kharbana already incorporates user intent in the query refinement process (Kharbana, ¶43). The proposed combination is incorporating the technqiues taught by RodriguezGarcia to use determine an expanded repository of known intents (RodriguezGarcia, ¶24) to use to refine the query (Kharbana, ¶43, ¶120).
With regard to claims 3 and 14 the proposed combination further teaches wherein the historical interaction information comprises:
a historical query statement as user search history (Kharbanda, ¶130 “The context determination block 62 may identify and/or process metadata, user profile data (e.g., preferences, user search history, user browsing history, user purchase history, and/or user input data), previous interaction data”) corresponding to the user identifier as the user profile associated with a particular user (Kharbanda, ¶130 “The context determination block 62 may identify and/or process metadata, user profile data … associated with the user. The context can be associated… associated with the user and/or the retrieved or obtained data.”) and at least one of:
a second candidate intent corresponding to the historical query statement as the known utterance-intent pair (RodriguezGarcia, ¶8 “In some implementations, the known intents are a subset of a plurality of known intents stored in a training dataset, where the training dataset includes a plurality of utterance-intent pairs of a particular domain.”) of the user search history (Kharbanda, ¶130);
second descriptive information (RodriguezGarcia, ¶52 ) corresponding to the second candidate intent as the known intent (RodriguezGarcia, Figure 2B, 214);
a second target intent corresponding to the historical query statement; or
a response statement corresponding to the historical query statement.
With regard to claim 5 RodriguezGarcia further teaches wherein obtaining the first candidate intent as the discovered intent (RodriguezGarcia, ¶53 “The prompt 400 may include instructions 402 and 406 instructing an LLM (e.g., the intent predictor 124) to discover intents for utterances using samples 404 as a guide”) corresponding to the query statement as for the utterance (RodriguezGarcia, ¶39 “here each utterance in both the few-shot pool 108 and the test batch”) by matching (RodriguezGarcia, ¶39 “based on a similarity measure (e.g., cosine distance) between the one or more examples and their respective test-batch utterance.”; ¶57 “If the utterance matches a known intent, assign it to that intent.”; ¶84 “which may use a pre-trained neural network model or another mechanism to determine whether an intent matching the utterance exists in a database 810 that stores known intents. If a match is found, the intent retriever 808 responds to the IVR module 806 with the matching intent.”; ¶94 “In some implementations, utterances corresponding to the list of known intents are semantically similar to utterances in the list of test examples. In some implementations, an utterance in the few-shot example is semantically similar to an utterance in at least one of the list of test examples.”) the query statement as the utterance in the test examples (Id) with the preset intent as the list of known intents (Id) and the preset descriptive information (RodriguezGarcia, ¶52 “the instructions 302 may be a task description that instructs a prompt generator (e.g., the prompt generator 106 described in FIG. 1) to generate a prompt for use by another LLM (e.g., the intent predictor 124 described in FIG. 1).”) of the preset intent as the known intents (RodriguezGarcia, ¶41) comprises:
determining a first as for the utterance (RodriguezGarcia, ¶39 “here each utterance in both the few-shot pool 108 and the test batch”) encoding vector (RodriguezGarcia, ¶39 “the test batch are embedded into vectors to enable the selection of one or more examples from the few-shot pool 108 for each test-batch utterance, based on a similarity measure (e.g., cosine distance) between the one or more examples and their respective test-batch utterance”) corresponding to the query statement as for the utterance (RodriguezGarcia, ¶39), and a second as the test batch (RodriguezGarcia, ¶39 “here each utterance in both the few-shot pool 108 and the test batch”) encoding vector (RodriguezGarcia, ¶39) corresponding to the preset intent as the known intents (RodriguezGarcia, ¶41 “the known intent feedback 120 is one or more utterance-intent pairs retrieved by the few-shot sampler 114 from known intents 112, which may be stored in a variety of storage formats, … . In some implementations, the known intents 112 may either be a copy of, or a subset of, the training dataset 104, augmented with one or more intents identified by the intent predictor 124.”) and the preset descriptive information (RodriguezGarcia, ¶52 “the instructions 302 may be a task description that instructs a prompt generator (e.g., the prompt generator 106 described in FIG. 1) to generate a prompt for use by another LLM (e.g., the intent predictor 124 described in FIG. 1).”) of the preset intent as the known intents (RodriguezGarcia, ¶41);
calculating a similarity (RodriguezGarcia, ¶39 “based on a similarity measure (e.g., cosine distance) between the one or more examples and their respective test-batch utterance.”; ¶94 “In some implementations, utterances corresponding to the list of known intents are semantically similar to utterances in the list of test examples. In some implementations, an utterance in the few-shot example is semantically similar to an utterance in at least one of the list of test examples.”) between the first encoding vector as the test examples (Id) and the second encoding vector as the list of known samples (Id);
and identifying a first number as the selection of one or more examples from the few-shot pool 108 (RodriguezGarcia, ¶39 “the selection of one or more examples from the few-shot pool 108 for each test-batch utterance, based on a similarity measure (e.g., cosine distance) between the one or more examples and their respective test-batch utterance”) of preset intents as from the few-shot pool 108, which is part of the known intent database (RodriguezGarcia, ¶38 “the few-shot sampler 114 may retrieve the one or more few-shot examples 116 from a few-shot pool 108, which includes a subset of the training dataset 104, for example, 10% of the samples for each known intent in the training dataset 104.”) with a [[ as most similar (RodriguezGarcia, ¶39) as the first candidate intent as the discovered intent (RodriguezGarcia, ¶53 “The prompt 400 may include instructions 402 and 406 instructing an LLM (e.g., the intent predictor 124) to discover intents for utterances using samples 404 as a guide”).
RodriguezGarcia does not explicitly teach highest similarity.
Kharbana teaches highest similarity (¶70 “The system can determine that the input audio signature is similar to a first known signature associated with a first known audio file when the similarity value exceeds a threshold value”).
It would have been obvious to one of ordinary skill to which said subject matter pertains at the time the invention was filed to have implemented the similarity taught by RodriguezGarcia to use the threshold evaluation when determining the most similar intents as it yields the predictable results of determining the most similar elements. Please note that one of ordinary skill in the art would recognize that the cosine distance calculation taught by RodriguezGarcia (¶39) may reasonably be expected to be evaluated using a threshold as taught by Kharbana (¶70) to determine the most similar vectors (RodriguezGarcia, ¶39).
With regard to claim 6 the proposed combination further teaches wherein determining the first as for the utterance (RodriguezGarcia, ¶39 “here each utterance in both the few-shot pool 108 and the test batch”) encoding vector (RodriguezGarcia, ¶39 “the test batch are embedded into vectors to enable the selection of one or more examples from the few-shot pool 108 for each test-batch utterance, based on a similarity measure (e.g., cosine distance) between the one or more examples and their respective test-batch utterance”) corresponding to the query statement as for the utterance (RodriguezGarcia, ¶39), and the second as the test batch (RodriguezGarcia, ¶39 “here each utterance in both the few-shot pool 108 and the test batch”) encoding vector (RodriguezGarcia, ¶39) corresponding to the preset intent as the known intents (RodriguezGarcia, ¶41 “the known intent feedback 120 is one or more utterance-intent pairs retrieved by the few-shot sampler 114 from known intents 112, which may be stored in a variety of storage formats, … . In some implementations, the known intents 112 may either be a copy of, or a subset of, the training dataset 104, augmented with one or more intents identified by the intent predictor 124.”) and the preset descriptive information (RodriguezGarcia, ¶52 “the instructions 302 may be a task description that instructs a prompt generator (e.g., the prompt generator 106 described in FIG. 1) to generate a prompt for use by another LLM (e.g., the intent predictor 124 described in FIG. 1).”) of the preset intent as the known intents (RodriguezGarcia, ¶41) comprises:
obtaining the first as for the utterance (RodriguezGarcia, ¶39 “here each utterance in both the few-shot pool 108 and the test batch”) encoding vector (RodriguezGarcia, ¶39 “the test batch are embedded into vectors to enable the selection of one or more examples from the few-shot pool 108 for each test-batch utterance, based on a similarity measure (e.g., cosine distance) between the one or more examples and their respective test-batch utterance”) by inputting the query statement into an intent retrieval model as the technique for selecting the subset of known intents 112 (RodriguezGarcia, ¶42 “the few-shot sampler 114 may use the KNN semantic sampling technique described earlier to select the subset of the known intents 112 that are semantically similar to the current test batch 118.”); and
obtaining the second as the test batch (RodriguezGarcia, ¶39 “here each utterance in both the few-shot pool 108 and the test batch”) encoding vector (RodriguezGarcia, ¶39) corresponding to the preset intent as the known intents (RodriguezGarcia, ¶41 “the known intent feedback 120 is one or more utterance-intent pairs retrieved by the few-shot sampler 114 from known intents 112, which may be stored in a variety of storage formats, … . In some implementations, the known intents 112 may either be a copy of, or a subset of, the training dataset 104, augmented with one or more intents identified by the intent predictor 124.”) by concatenating (RodriguezGarcia, ¶40 “the few-shot examples 116 then may be concatenated with the samples selected from the training dataset 104 to constitute a sequence of samples to be fed to the intent predictor 124.”) and inputting the preset intent as the training dataset (Id) and the descriptive information (RodriguezGarcia, ¶52 “the instructions 302 may be a task description that instructs a prompt generator (e.g., the prompt generator 106 described in FIG. 1) to generate a prompt for use by another LLM (e.g., the intent predictor 124 described in FIG. 1).”) corresponding to the preset intent into the intent retrieval model (RodriguezGarcia, ¶50 “In some implementations, the one or more software components additionally incorporate into the augmented prompt 216 one or more utterances from test examples 210, which are utterances whose intents are to be discovered by an LLM 218 (e.g., the intent predictor 124 described in FIG. 1).”),
wherein the intent retrieval model is generated by training a pre-trained language model (RodriguezGarcia, ¶84 “the IVR module 806 may send the utterance to an intent retriever 808, which may use a pre-trained neural network model or another mechanism to determine whether an intent matching the utterance exists in a database 810 that stores known intents”) based on a sample statement as the utterance (Id) and a corresponding intent label as the matching intent (Id).
Response to Arguments
Applicant's arguments filed June 13, 2026 have been fully considered but they are not persuasive.
With regard to the abstract, the abstract submitted June 13, 2026 is not a paragraph, and is instead a single sentence written in claim form. It is suggested that the abstract be amended to be a paragraph, and written in a narrative format.
With regard to the 101 rejection of claim 19, the rejection is hereby withdrawn in view of the claim amendments.
Response to Arguments
Applicant's arguments filed July 13, 2026 have been fully considered but they are not persuasive.
With regard to claim “Distinguishing Feature 1”: applicant argues that the claim language requires that the first candidate intent is selected exclusively from within the existing pool of preset intents. Applicant argues that RodriguezGarcia discovers intents that are NOT in the known intent list, citing Paragraph [0056] “The intent can be one of the pre-defined intents or a new one that you create based on the context and knowledge about the problem and specific data domain”.
In response to applicant's argument that the references fail to show certain features of the invention, it is noted that the features upon which applicant relies (i.e., the intent is selected exclusively from within the existing pool of preset intents) are not recited in the rejected claim(s). Although the claims are interpreted in light of the specification, limitations from the specification are not read into the claims. See In re Van Geuns, 988 F.2d 1181, 26 USPQ2d 1057 (Fed. Cir. 1993). The language does require the situation where the first candidate intent is obtained from the preset intents, but this does not require an exclusive selection. The language details how the obtaining is determined, “by matching the query statement with the preset intent”. As long as there is a first candidate intent that is selected from the preset intent by matching the query statement with the preset intent, this does not require that all intents be exclusively from the preset intent. RodriguezGarcia teaches using a similarity measure to perform the intent selection, and acknowledge that the selected intent may be selected from the known intents. Applicant acknowledges in the cited Paragraph [0056] which explicitly details that “The intent can be one of the pre-defined intents.” Alternative, or additional operations do not invalidate the recitation that the intent can be one of the pre-defined intents.
Applicant argues that the training samples used for the KNN (Citing to RodriguezGarcia [0039]) are not candidate intents selected from within known intent sets.
In response, one of ordinary skill in the art would readily recognize the training samples as including known intents. Furthermore, the citation to Paragraph [0039] was provided to address how matching is performed (e.g. cosine distance similarity). Applicant’s arguments do not address the claim mapping. More specific citations were provided where RodriguezGarcia recites discovering intent by matching the query statement with known intents.
Applicant argues that RodriguezGarcia “discovered intents’ may be enterily novel intents that are created by the LLM, while the claim constrains the output to intents that already exist among the preset intents.
In response, the instant claim language does not constrain the output intents to only already existing preset intents. The claim language has a positive recitation of an intent that is selected from the already existing preset intents. RodriguezGarcia explicitly recites situations were the intent is selected from known intents, and thus reads on the claim limitations. RodriguezGarcia indicates that known intents are reused as much as possible ([0058]) and the system only creates new intents when the matching to known intents has failed ([0084]). The instant claim language only recites the situation where a match occurs, and is silent regarding what occurs when no match is found.
With regard to “Distinguishing Feature (2)”: Applicant argues a specific multi step process that is not required by the claim language. Applicant argues that the iterations taught by RodriguezGarcia are not done within a single query statement, and each iteration is performed using a different training set.
In response, RodriguezGarcia does not require each iteration to be performed with an entirely different training set. The data from previous iterations are used for subsequent iterations (Paragraph [0070]). Furthermore, applicant’s arguments do not address the claim mapping. The first prompt was mapped to the initial prompt 208 that is generated (See paragraph [0048]). The system then uses this prompt to discover an intent and generate an augments prompt 216 which was mapped to the second prompt information (See paragraph [0050]). The system then uses the augmented prompt as input to discover a second intent (See Paragraph [0051]). The second intent is not selected from the Few-shot pool as the first one was, but is instead selected the known intents 214. This is all performed in a single iteration of the device. Applicant’s arguments regarding the iteration of the device is irrelevant. The iteration of the device does not occur until the system updates the known intents (Figure 2, 222), which was mapped to the generating new second prompt information, as this is the information that will be used to generate the next prompt. Applicant argues generally against the prior art and does not appear to address the claim mapping put forth.
Conclusion
THIS ACTION IS MADE FINAL. Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to AMANDA WILLIS whose telephone number is (571)270-7691. The examiner can normally be reached Monday-Friday 8am-2pm.
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, Ajay Bhatia can be reached at 571-272-3906. 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.
/AMANDA L WILLIS/ Primary Examiner, Art Unit 2156
Prosecution Insights