DETAILED ACTION
Notice of 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 .
Claim Objections
Claim 15 is objected to because of the following informalities:
Line 1 “wherein the list of known intents include at one intent from a training dataset” should be changed to “ wherein the list of known intents include at least one intent from a training dataset “.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 1-20 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an
abstract idea without significantly more.
The Independent claims 1, 9 and 16 recite “obtaining an utterance and a corresponding label representative of an intent of the utterance”; “generating, using a first large language model on the utterance and the corresponding label, a prompt comprising a task description and an input-label pair”; “modifying the prompt to include one or more of: a list of known intents”; “a few-shot example”; “or a list of test examples”; “and generating, using a second large language model on the modified prompt, a list of predicted intents”; “determining that a particular intent in the list of predicted intents is not in the list of known intents”; “and updating the list of known intents with the particular intent”. The limitations above as drafted, is a process that, under its broadest reasonable interpretation, covers a mental process, as this could be performed in the human mind or with the aid of pen and paper.
The limitation of " obtaining ... ", "generating ... ", "modifying ... ", “determining..”, “updating…” as drafted covers mental activities. More specifically, a human can obtain an utterance related to an intent during the conversation with another human being, such as travel plan. One can generate a response based on the intent and general knowledge of the subject matter. Responses can be modified based on known intent or know responses or less examples ( few shot example), such as preferred travel dates, companion etc. by using more specific knowledge regarding the subject matter. During the modification, it can be find out that their might be some intent that was unknown such as packing some food items. That unknown intent can be added to the list of intents. All the steps above are examples of observation and evaluation that could be performed in the human mind or with the aid of pencil and paper.
The claims recite the additional limitation of “large language model”, claims 9 and 16 recite “processor”, claim 9 recites “memory” and claim 16 recites “ non transitory computer readable storage medium” for performing the method. All those are recited at a high level of generality and are recited as performing generic computer functions routinely used in computer applications. The current specification in paragraph [0030] specifies large language model as neural network or machine learning model, which is generic and can be any neural network or any machine learning model, which is not sufficient to amount to significantly more than the judicial exception. Specification in paragraphs [0102],[0109], clearly specifies “processor”, “memory” and “ non transitory computer readable storage medium” as performing generic computer functions that are well-understood, routine and conventional activities amount to no more than implementing the abstract idea with a computerized system. The claims as drafted, are not patent eligible
Thus, taken alone, the additional elements do not amount to significantly more than the above identified judicial exception (the abstract idea). Looking at the limitations as an ordered combination adds
nothing that is not already present when looking at the elements taken individually. There is no indication
that the combination of elements improves the functioning of a computer or improves any other technology. Their collective functions merely provide conventional computer implementation. Claims 1, 9
and 16 are therefore not drawn to eligible subject matter as they are directed to an abstract idea without
significantly more than the abstract idea.
Claim 2 recites the additional limitation of “wherein the utterance and the corresponding label are retrieved from a training dataset that includes a plurality of utterance-intent pairs for a particular domain” , where extracting/using a number of utterance-intent pair of a particular area, from a written documents of utterance -intent pair, could be performed in the human mind or with the aid of pen and paper. The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception, as claim 2 does not recite any additional limitations. The claim as drafted, is not patent eligible.
Claim 3 recites “wherein utterances corresponding to the list of known intents are semantically similar to utterances in the list of test examples”, to find out that the utterances of one list is semantically similar to another list is an evaluation, observation and could be performed in the human mind or with the aid of pen and paper. The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception, as claim 3 does not recite any additional limitations. The claim as drafted, is not patent eligible.
Claim 4 recites “wherein an utterance in the few-shot example is semantically similar to an utterance in at least one of the list of test examples”, to find out that the utterances of one list, which is smaller in size ( few shot) is semantically similar to another list is an evaluation, observation and could be performed in the human mind or with the aid of pen and paper. The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception, as claim 4 does not recite any additional limitations. The claim as drafted, is not patent eligible.
Claim 5 recites “wherein each of the first large language model and the second large language model is a frozen transformer”, where first and second language model are additional elements which are specified in specification in paragraph [0030] as neural network or machine learning model and “frozen transformer”, is one of the characteristics of the language model, which is generic and can be any neural network or any machine learning model, which are not sufficient to amount to significantly more than the judicial exception. The claim 5 as drafted, is not patent eligible.
Claims 6, 11 recite “wherein utterances in the list of known intents and utterances in the few-shot example are of a same domain”, to find out that the utterances of two lists are from the same domain is an evaluation, observation and could be performed in the human mind or with the aid of pen and paper. The claims do not include additional elements that are sufficient to amount to significantly more than the judicial exception, as claims 6 and 11 do not recite any additional limitations. The claims as drafted, are not patent eligible.
Claim 7 recites “wherein the first large language model is to generate the prompt based on a template”, where generating some prompt or input based on some instruction ( template) could be performed with the aid of pen and paper. The claim recites additional limitation of language model, which is specified in specification in paragraph [0030] as neural network or machine learning model, which is generic and can be any neural network or any machine learning model, which is not sufficient to amount to significantly more than the judicial exception. The claim 7 as drafted, is not patent eligible.
Claim 8 recites “wherein the list of test examples includes a plurality of utterances without matching intents, and where at least one of the plurality of utterances is received from a caller via a server in a call center”, where determining that the list of some of the test utterances doesn’t have matching intent and determining some of the utterances are from call center could be an evaluation, observation and could be performed in the human mind or with the aid of pen and paper. The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception, as claim 8 does not recite any additional limitations. The claim as drafted, is not patent eligible.
Claim 10 recites “wherein the utterance and the corresponding label is randomly selected from a training dataset, wherein the training dataset includes a plurality of utterance-intent pairs of a particular domain”, where extracting/using a number of utterance-intent pair of a particular area, from a written documents of utterance -intent pair, could be performed in the human mind or with the aid of pen and paper. The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception, as claim 10 does not recite any additional limitations. The claim as drafted, is not patent eligible.
Claim 12 recites “wherein the first large language model is to generate the prompt based on a template that specifies a format of a response that the second large language model is to return”, determining that the prompt is based on a format is an evaluation, observation and could be performed in the human mind or with the aid of pen and paper. The claim recites additional limitation of language model, which is specified in specification in paragraph [0030] as neural network or machine learning model, which is generic and can be any neural network or any machine learning model, which is not sufficient to amount to significantly more than the judicial exception. The claim 12 as drafted, is not patent eligible.
Claim 13 recites “wherein at least one of a plurality of utterances in the list of test examples is received from a caller via a server in a call center, and wherein the server is to generate a response to the caller using an intent returned by the second large language model based on the at least one utterance”, to determine that the utterances of one list are from call center and generating response based on the intent could be an evaluation, observation and could be performed in the human mind or with the aid of pen and paper. The claim recites additional limitations of language model and server. Language model is specified in specification in paragraph [0030] as neural network or machine learning model, which is generic and can be any neural network or any machine learning model, which is not sufficient to amount to significantly more than the judicial exception. Server is specified in specification, para.[0087]-[0090] as performing generic computer functions, which is not sufficient to amount to significantly more than the judicial exception. The claim 13 as drafted, is not patent eligible.
Claim 14 recites “wherein the prompt includes a place holder for the few-shot example to be inserted into the prompt, and wherein the prompt further includes one or more instructions instructing the second large language model how to use the few-shot example in discovering intents for the list of test examples”, to find out that the prompt has a certain place for some example which are fewer in number ( few shot example) and to determine that includes some instruction for the language model, are an evaluation, observation and could be performed in the human mind or with the aid of pen and paper. The claim recites additional limitation of language model. Language model is specified in specification in paragraph [0030] as neural network or machine learning model, which is generic and can be any neural network or any machine learning model, which is not sufficient to amount to significantly more than the judicial exception. The claim 14 as drafted, is not patent eligible.
Claim 15 recites “wherein the list of known intents include at one intent from a training dataset and at least one intent discovered by the second large language model in a previous iteration”, to find out the source of the intents in intent list is an evaluation, observation and could be performed in the human mind or with the aid of pen and paper. The claim recites additional limitation of language model. Language model is specified in specification in paragraph [0030] as neural network or machine learning model, which is generic and can be any neural network or any machine learning model, which is not sufficient to amount to significantly more than the judicial exception. The claim 15 as drafted, is not patent eligible.
Claim 17 recites “wherein the known intents are a subset of a plurality of known intents stored in a training dataset, wherein the training dataset includes a plurality of utterance-intent pairs of a particular domain”, where human can create a dataset of utterance-intent pairs based on particular area of interest, which can be treated as a training dataset, to find out the area/domain of the utterance intent pair from the dataset is an evaluation, observation and could be performed in the human mind or with the aid of pen and paper. The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception, as claim 17 does not recite any additional limitations. The claim as drafted, is not patent eligible.
Claim 18 recites “wherein the few-shot example is selected from a few-shot pool that includes a subset of a training dataset, wherein the training dataset includes a plurality of utterance-intent pairs of a particular domain”, where human can create a dataset of utterance-intent pairs ( few shot examples) based on particular area of interest, which can be treated as a training dataset, to determine the source of the few shot examples and the domain of utterance intent pair is an evaluation, observation and could be performed in the human mind or with the aid of pen and paper. The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception, as claim 18 does not recite any additional limitations. The claim as drafted, is not patent eligible.
Claim 19 recites “wherein the updated list of known intents are to be inserted into the prompt”, using the updated intent as an input ( prompt) to determine the utterance intent could be performed in the human mind or with the aid of pen and paper. The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception, as claim 19 does not recite any additional limitations. The claim as drafted, is not patent eligible.
Claim 20 recites “wherein a server in a call center is to obtain the updated list of known intents and is to generate a response to a caller based on the updated list of known intents and an utterance of the caller.”, to determine that a call center received an updated list of intents is an evaluation, observation and could be performed in the human mind or with the aid of pen and paper. The claim recites additional limitations of server. Server is specified in specification, para.[0087]-[0090] as performing generic computer functions, which is not sufficient to amount to significantly more than the judicial exception. The claim 20 as drafted, is not patent eligible.
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 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 of this title, 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 1-4, 6-9, 11-13 and 15-17 are rejected under 35 U.S.C. 103 as being unpatentable over Mehrotra et al. ( US 20240143945 A1), hereinafter referenced as Mehrotra, in view of Levi et al. ( Intent-based Prompt Calibration: Enhancing prompt optimization with synthetic boundary cases, 04 March, 2024, ICLR 2024 Workshop), hereinafter referenced as Levi, further in view of Reddy et al. ( US 20250086394 A1), hereinafter referenced as Reddy.
Regarding Claim 1, Mehrotra teaches a method comprising:
obtaining an utterance and a corresponding label representative of an intent of the utterance ( Mehrotra: Para.[0056], Figs. 5, 2A, utterances annotated with intent labels ( labeled data 202) are obtained );
determining that a particular intent in the list of predicted intents is not in the list of known intents (Mehrotra: Para.[0026], Fig. 2B illustrates an example structure of the intent classification framework 250 with OOD identification which triggers out of domain (OOD) detection when a user enters a sentence that does not belong to any of the pre-defined intent classes) ;
Mehrotra while teaching the method of claim 1, fails to explicitly teach the claimed, generating, using a first large language model on the utterance and the corresponding label, a prompt comprising a task description and an input-label pair ; modifying the prompt to include one or more of: a list of known intents; a few-shot example; or a list of test examples; and generating, using a second large language model on the modified prompt, a list of predicted intents; and updating the list of known intents with the particular intent.
However, Levi does teach the claimed, generating, using a first large language model on the utterance and the corresponding label, a prompt comprising a task description and an input-label pair ( Levi: Section 1, paragraph 4, section 2, paragraph 1, 2, Figs. 1,2, Intent based prompt calibration system (IPC) generates calibrated prompt comprised of a task description calibrated to user’s intent, by using large language model based on user’s initial prompt ( input utterance) with a task description );
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate Levi’s teaching of intent based prompt calibration using large language model, into the system and method of intent classification to provide natural language understanding in a conversation agent , taught by Mehrotra, because, automatic prompt engineering would help to achieve optimized performance from LLMs. (Levi[ Section 5]).
Mehrotra in view of Levi while teaching the method of claim 1, fails to explicitly teach the claimed, modifying the prompt to include one or more of: a list of known intents; a few-shot example; or a list of test examples; and generating, using a second large language model on the modified prompt, a list of predicted intents; and updating the list of known intents with the particular intent.
However, Reddy does teach the claimed, modifying the prompt to include one or more of: a list of known intents; a few-shot example; or a list of test examples ( Reddy: Para.[0080]-[0082], prompts can be modified/improved by few-shot prompts shown in the table) ;
and generating, using a second large language model on the modified prompt, a list of predicted intents ( Reddy: Para.[0105]-[0108], Fig. 4, system 400 illustrates generating a digital assistant definition with one or more large language models ( first and second language models). Second Language model 450 generates list of intents from the prompt 460, which is from exemplar utterance ( enhanced prompt)) ;
and updating the list of known intents with the particular intent ( Reddy: Para.[ 0169], updating intents).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate Reddy’s teaching of digital assistant generation via large language models, into the system and method, taught by Mehrotra in view of Levi, because, by leveraging large language models to generate user operations or intents ( e.g., along with named entities) would develop high quality digital assistants. (Reddy[ Para. [0018]-[0021]).
Claim 9 is system claim comprising: one or more processors; and memory, including computer-executable instructions that, when executed by the one or more processor ( Mehrotra: Para.[0032], Fig. 3, processor 310 coupled to memory 320), cause the system to perform the steps in method claim 1 above and as such, claim 9 is similar in scope and content to claim 1 and therefore, claim 9 is rejected under similar rationale as presented against claim 1 above.
Claim 16 is non-transitory computer-readable storage medium claim having stored thereon executable instructions which, when executed by one or more processor of a computer system ( Mehrotra: Para. [0035], Fig. 3, memory 320 may include non-transitory, tangible, machine readable media that includes executable code that when run by one or more processors (e.g., processor 310) may cause the one or more processors to perform the methods described) , cause the computer system to perform the steps in method claim 1 above and as such, claim 16 is similar in scope and content to claim 1 and therefore, claim 16 is rejected under similar rationale as presented against claim 1 above.
Regarding Claim 2, Mehrotra in view of Levi, further in view of Reddy teach the method of claim 1. Mehrotra further teaches, wherein the utterance and the corresponding label are retrieved from a training dataset that includes a plurality of utterance-intent pairs for a particular domain ( Mehrotra: Para.[0056], Figs. 5, 2A, utterances annotated with intent labels ( labeled data 202) are obtained from the training dataset. Para.[0018], training dataset could be from different domains such as booking, healthcare, information technology support, banking customer service );
Regarding Claim 3, Mehrotra in view of Levi, further in view of Reddy teach the method of claim 1. Mehrotra further teaches, wherein utterances corresponding to the list of known intents are semantically similar to utterances in the list of test examples ( Mehrotra: Para.[0060], Fig.5, at step 510 semantic similarity between the testing utterance and training utterance ( utterance with known intent)).
Regarding Claim 4, Mehrotra in view of Levi, further in view of Reddy teach the method of claim 1. Reddy further teaches, wherein an utterance in the few-shot example is semantically similar to an utterance in at least one of the list of test examples (Reddy: Para.[0068], utterance of known intent of getting order status. Para.[0082], table illustrates few shot example of getting order details which is semantically similar to the test utterances).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate Reddy’s teaching of digital assistant generation via large language models, into the system and method of intent classification to provide natural language understanding in a conversation agent , taught by Mehrotra, because, by leveraging large language models to generate user operations or intents ( e.g., along with named entities) would develop high quality digital assistants. (Reddy[ Para. [0018]-[0021]).
Regarding Claim 6, Mehrotra in view of Levi, further in view of Reddy teach the method of claim 1. Reddy further teaches, wherein utterances in the list of known intents and utterances in the few-shot example are of a same domain (Reddy: Para.[0068], utterance of known intent of getting order status. Para.[0082], table illustrates few shot example of getting order details which is same as the utterance of known intent and from same domain).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate Reddy’s teaching of digital assistant generation via large language models, into the system and method of intent classification to provide natural language understanding in a conversation agent , taught by Mehrotra, because, by leveraging large language models to generate user operations or intents ( e.g., along with named entities) would develop high quality digital assistants. (Reddy[ Para. [0018]-[0021]).
Claim 11 is system claim performing the steps in method claim 3 above and as such, claim 11 is similar in scope and content to claim 3 and therefore, claim 11 is rejected under similar rationale as presented against claim 3 above.
Regarding Claim 7, Mehrotra in view of Levi, further in view of Reddy teach the method of claim 1. Reddy further teaches, wherein the first large language model is to generate the prompt based on a template ( Reddy: Para.[0076], prompts can be generated based on template. Para.[0105],[0106], Fig.4 illustrates a system 400 with multiple language models, where large language models 420A, 420B are the first LLM).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate Reddy’s teaching of digital assistant generation via large language models, into the system and method, taught by Mehrotra in view of Levi, because, by leveraging large language models to generate user operations or intents ( e.g., along with named entities) would develop high quality digital assistants. (Reddy[ Para. [0018]-[0021]).
Regarding Claim 8, Mehrotra in view of Levi, further in view of Reddy teach the method of claim 1, wherein the list of test examples includes a plurality of utterances without matching intents ( Mehrotra: Para.[0069], Fig. 6, at step 608 testing utterance is determined to be OOD which means no matching intent),
and where at least one of the plurality of utterances is received from a caller via a server in a call center ( Mehrotra: Para.[0058], testing utterance could be received via a communication interface. Para.[0018], large training/testing dataset could be from different domains such as healthcare, information technology support, banking customer service).
Regarding Claim 12, Mehrotra in view of Levi, further in view of Reddy teach the system of claim 9. Reddy further teaches, wherein the first large language model is to generate the prompt based on a template that specifies a format of a response that the second large language model is to return ( Reddy: Para.[0105], Fig. 4, illustrates a system 400 with multiple language models, where first LLMs are 420A, 420B and second LLM is 450. Para.[0076], prompts can be generated based on template. Para.[0080], Prompts can include examples to encourage the large language model to provide results in a desired format).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate Reddy’s teaching of digital assistant generation via large language models, into the system and method of intent classification to provide natural language understanding in a conversation agent , taught by Mehrotra in view of Levi, because, by leveraging large language models to generate user operations or intents ( e.g., along with named entities) would develop high quality digital assistants. (Reddy[ Para. [0018]-[0021]).
Regarding Claim 13, Mehrotra in view of Levi, further in view of Reddy teach the system of claim 9. Mehrotra further teaches, wherein at least one of a plurality of utterances in the list of test examples is received from a caller via a server in a call center ( Mehrotra: Para.[0058], testing utterance could be received via a communication interface. Para.[0018], large training/testing dataset could be conversations from different domains such as healthcare, information technology support, banking customer service),
and wherein the server is to generate a response to the caller using an intent returned by the second large language model based on the at least one utterance (Mehrotra: Para.[0043], Fig. 4, the user device 410 may receive a message indicating an intent classification label, or a system response generated based on the intent classification label, from the server 430 and display the message via the UI application 412. Para.[0020], Fig. 2A, shows an intent classification framework 200 which uses multilingual embeddings 222 which may be language-agnostic BERT sentence representations (LaBSE) (second large language model)).
Regarding Claim 15, Mehrotra in view of Levi, further in view of Reddy teach the system of claim 9. Mehrotra further teaches, wherein the list of known intents include at one intent from a training dataset and at least one intent discovered by the second large language model in a previous iteration (Mehrotra: Para.[0047],[0049],[0050], the utterance intent pairs can contain data from training dataset and from the intent classification module).
Regarding Claim 17, Mehrotra in view of Levi, further in view of Reddy teach the non-transitory computer-readable storage medium of claim 16. Mehrotra further teaches, wherein the known intents are a subset of a plurality of known intents stored in a training dataset, wherein the training dataset includes a plurality of utterance-intent pairs of a particular domain ( Mehrotra: Para.[0047], datasets of utterance-intent pairs are from training datasets. Para.[0018], training dataset could be conversations from different domains such as healthcare, information technology support, banking customer service ( particular domain)).
Claim 5 is rejected under 35 U.S.C. 103 as being unpatentable over Mehrotra et al. ( US 20240143945 A1), hereinafter referenced as Mehrotra, in view of Levi et al. ( Intent-based Prompt Calibration: Enhancing prompt optimization with synthetic boundary cases, 04 March, 2024, ICLR 2024 Workshop), hereinafter referenced as Levi, in view of Reddy et al. ( US 20250086394 A1), hereinafter referenced as Reddy, further in view of Chuang et al. ( A Soft Prompt-Based Calibration on Mitigating Performance Variability in Clinical Notes Summarization" Rice University, Houston, TX; University of Texas Health Science Center, Houston, TX, 2023), hereinafter referenced as Chuang.
Regarding Claim 5, Mehrotra in view of Levi, further in view of Reddy teach the method of claim 1. Mehrotra in view of Levi, further in view of Reddy fail to explicitly teach the claimed, further teaches, wherein each of the first large language model and the second large language model is a frozen transformer.
However, Chuang does teach the claimed, further teaches, wherein each of the first large language model and the second large language model is a frozen transformer ( Chuang: Section 4, paragraph 1, Fig. 1 illustrates two large language models with frozen transformer).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate Chuang’s teaching of soft prompt based calibration pipeline to generate summarization, into the system and method, taught by Mehrotra in view of Levi, further in view of Reddy , because, by using large language models and soft prompts in conjunction with discrete prompts, generation of summarizing of clinical notes would be improved. (Chuang, [ Section 6]).
Claim 10 is rejected under 35 U.S.C. 103 as being unpatentable over Mehrotra et al. ( US 20240143945 A1), hereinafter referenced as Mehrotra, in view of Levi et al. ( Intent-based Prompt Calibration: Enhancing prompt optimization with synthetic boundary cases, 04 March, 2024, ICLR 2024 Workshop), hereinafter referenced as Levi, in view of Reddy et al. ( US 20250086394 A1), hereinafter referenced as Reddy, further in view of Xia et al. ( US 20210374603 A1), hereinafter referenced as Xia.
Regarding Claim 10, Mehrotra in view of Levi, further in view of Reddy teach the system of claim 9. Mehrotra in view of Levi, further in view of Reddy fail to explicitly teach, wherein the utterance and the corresponding label is randomly selected from a training dataset, wherein the training dataset includes a plurality of utterance-intent pairs of a particular domain.
However, Xia does teach the claimed, wherein the utterance and the corresponding label is randomly selected from a training dataset, wherein the training dataset includes a plurality of utterance-intent pairs of a particular domain ( Xia: Para.[0083], Fig. 10 provides an example data table showing the dataset details of SNIPS-NLU and NLUED. For example, SNIPS-NLU contains seven intents in total. Two of them are regraded as few-shot intents. The others are used as existing intents with sufficient annotation. 80% of the whole data is randomly chosen as the training data and 20% as the test data. NLUED2 is a natural language understanding dataset with 64 intents for human-robot interaction in home domain, in which 16 intents as randomly selected as the few-shot ones. A sub-corpus of 11,036 utterances with 10-folds cross-validation splits is utilized).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate Xia’s teaching of composed variational natural language generation (CLANG) model that is configured to generate training samples for few-shot intents, into the system and method, taught by Mehrotra in view of Levi, further in view of Reddy, because, by providing a balanced training dataset, user intentions from a user spoken input can be identified effectively to generate an effective natural language response from an intelligent assistant. (Xia[ Para. [0003],[0023]).
Claims 14, 18 are rejected under 35 U.S.C. 103 as being unpatentable over Mehrotra et al. ( US 20240143945 A1), hereinafter referenced as Mehrotra, in view of Levi et al. ( Intent-based Prompt Calibration: Enhancing prompt optimization with synthetic boundary cases, 04 March, 2024, ICLR 2024 Workshop), hereinafter referenced as Levi, in view of Reddy et al. ( US 20250086394 A1), hereinafter referenced as Reddy, further in view of Bose et al. ( US 20240379096 A1), hereinafter referenced as Bose.
Regarding Claim 14, Mehrotra in view of Levi, further in view of Reddy teach the system of claim 9. Mehrotra in view of Levi, further in view of Reddy fail to explicitly teach, further teaches, wherein the prompt includes a place holder for the few-shot example to be inserted into the prompt, and wherein the prompt further includes one or more instructions instructing the second large language model how to use the few-shot example in discovering intents for the list of test examples.
However, Bose does teach the claimed, further teaches, wherein the prompt includes a place holder for the few-shot example to be inserted into the prompt ( Bose: Para.[0026]- [0030], Fig. 3 illustrates prompt which has place holder/slot values for few-shot example such as "<utterance>" -> "<intent> (value1, value2, ... )" ),
and wherein the prompt further includes one or more instructions instructing the second large language model how to use the few-shot example in discovering intents for the list of test examples ( Bose: Para.[0061], Fig. 6, at step 606 The prompt generator 110 creates the prompt which includes instructions to the model for the target task, the few-shot examples, and the utterance. At 608, prompt is input into the large language model and at 610, intent is obtained from the large language model ).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate Bose’s teaching of augmenting a prompt with a few-shot examples for a large language model to predict an intent given an utterance, into the system and method, taught by Mehrotra in view of Levi, further in view of Reddy , because, by using low cost solution of few-shot prompting, an effective intent prediction model can be achieved. (Bose, Para.[0014]-[0017]).
Regarding Claim 18, Mehrotra in view of Levi, further in view of Reddy teach the system of claim 9. Mehrotra in view of Levi, further in view of Reddy fail to explicitly teach, wherein the few-shot example is selected from a few-shot pool that includes a subset of a training dataset, wherein the training dataset includes a plurality of utterance-intent pairs of a particular domain.
However, Bose does teach the claimed, wherein the few-shot example is selected from a few-shot pool that includes a subset of a training dataset, [wherein the training dataset includes a plurality of utterance-intent pairs of a particular domain] ( Bose: Para.[0022],[0057], few shot examples are extracted from database 116 of utterance-intent ( pool of few shot examples)),
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate Bose’s teaching of augmenting a prompt with a few-shot examples for a large language model to predict an intent given an utterance, into the system and method, taught by Mehrotra in view of Levi, further in view of Reddy , because, by using low cost solution of few-shot prompting, an effective intent prediction model can be achieved. (Bose, Para.[0014]-[0017]).
Mehrotra further teaches, wherein the training dataset includes a plurality of utterance-intent pairs of a particular domain( Mehrotra: Para.[0018], training dataset could be from different domains such as booking, healthcare, information technology support, banking customer service );
Claim 19 is rejected under 35 U.S.C. 103 as being unpatentable over Mehrotra et al. ( US 20240143945 A1), hereinafter referenced as Mehrotra, in view of Levi et al. ( Intent-based Prompt Calibration: Enhancing prompt optimization with synthetic boundary cases, 04 March, 2024, ICLR 2024 Workshop), hereinafter referenced as Levi, in view of Reddy et al. ( US 20250086394 A1), hereinafter referenced as Reddy, further in view of Brockett et al. ( US 20230405468 A1), hereinafter referenced as Brockett.
Regarding Claim 19, Mehrotra in view of Levi, further in view of Reddy teach the non-transitory computer-readable storage medium of claim 16. Mehrotra in view of Levi, further in view of Reddy fail to explicitly teach, wherein the updated list of known intents are to be inserted into the prompt.
However, Brockett does teach the claimed, wherein the updated list of known intents are to be inserted into the prompt ( Brockett: Para.[0036], user’s intent may change overtime and based on that prompt is adjusted ).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate Brockett’s teaching of systems and methods which utilizes machine learning techniques to provide enhanced accessibility features to a game, into the system and method, taught by Mehrotra in view of Levi, further in view of Reddy , because, a robust system can be developed with accessibility to features which can be customized to allow users having different gameplay needs to enjoy the benefits of a game. (Brockett, Para.[0004]-[0006]).
Claim 20 is rejected under 35 U.S.C. 103 as being unpatentable over Mehrotra et al. ( US 20240143945 A1), hereinafter referenced as Mehrotra, in view of Levi et al. ( Intent-based Prompt Calibration: Enhancing prompt optimization with synthetic boundary cases, 04 March, 2024, ICLR 2024 Workshop), hereinafter referenced as Levi, in view of Reddy et al. ( US 20250086394 A1), hereinafter referenced as Reddy, further in view of Choudhary et al. ( US 20210224818 A1), hereinafter referenced as Choudhary.
Regarding Claim 20, Mehrotra in view of Levi, further in view of Reddy teach the non-transitory computer-readable storage medium of claim 16. Mehrotra in view of Levi, further in view of Reddy fail to explicitly teach the claimed, wherein a server in a call center is to obtain the updated list of known intents and is to generate a response to a caller based on the updated list of known intents and an utterance of the caller.
However, Choudhary does teach the claimed, wherein a server in a call center is to obtain the updated list of known intents (Choudhary: Para.[0043]-[0047], the customer service receives updated intent list and can display the updated or revised list of intents in the user interface. Para.[0178], Fig. 6, server 622)
and is to generate a response to a caller based on the updated list of known intents and an utterance of the caller (Choudhary: Para.[0048]-[0050], based on the updated intent list, an agent can respond to caller’s request ).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate Choudhary’s teaching of process flow for implementing a conversation-based customer service support system, into the system and method, taught by Mehrotra in view of Levi, further in view of Reddy , because, this would improve the accuracy of the customer service system. (Choudhary, Para.[0004]-[0013]).
Conclusion
Listed below are the prior arts made of record and not relied upon but are considered pertinent to applicant's disclosure.
Vishnoi et al. (US 11651768 B2) teaches techniques for stop word data augmentation for training chatbot systems in natural language processing. In one particular aspect, a computer-implemented method includes receiving a training set of utterances for training an intent classifier to identify one or more intents for one or more utterances; augmenting the training set of utterances with stop words to generate an augmented training set of out-of-domain utterances for an unresolved intent category corresponding to an unresolved intent; and training the intent classifier using the training set of utterances and the augmented training set of out-of-domain utterances. The augmenting includes: selecting one or more utterances from the training set of utterances, and for each selected utterance, preserving existing stop words within the utterance and replacing at least one non-stop word within the utterance with a stop word or stop word phrase selected from a list of stop words to generate an out-of-domain utterance.
Weston et al. (US 20250054493 A1) teaches an electronic device which includes a memory and a processor. The processor is configured to provide an utterance to an NLU model, which selects an intent from the list of intents based on the utterance. The processor is also configured to cause a response to the utterance to be provided to a user based on the selected intent. The NLU may be trained by obtaining a dataset comprising one or more words, generating a list of intents based at least in part on the dataset, grouping the list of intents into one or more domains, generating a list of training utterances for each intent in the list of intents, and modifying one or more parameters of the NLU model based on the list of intents, the list of training utterances, and/or the one or more domains.
Subramanya et al. (US 11687802 B2) teaches systems and methods for proactively predicting user intents on personal agents. In some examples, a user intent prediction system can include a computing device configured to obtain user intent data identifying a desired action by a user on a network-enabled tool. The computing device is further configured to obtain contextual data characterizing a user's interaction with the network-enabled tool. The computing device can then determine at least one predicted future intent of the user on the network-enabled tool based on the user intent data and the contextual data and present the at least one predicted future intent to the user.
Fu et al. (US 20250103325 A1 ) teaches a code review which is automatically generated by a large language model given a prompt that includes code changes made to a source code program, an associated intent, and an extended context. The intent represents an issue with the code changes from a code reviewer's perspective and is predicted from a neural classifier given the code changes in a code diff format. The neural classifier is a neural encoder transformer model pre-trained on various code review datasets and fine-tuned on code diff hunks of code changes labeled with an intent.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to NADIRA SULTANA whose telephone number is (571)272-4048. The examiner can normally be reached M-F,7:30 am-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, Paras D. Shah can be reached on (571) 270-1650. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/NADIRA SULTANA/Examiner, Art Unit 2653
/Paras D Shah/Supervisory Patent Examiner, Art Unit 2653
11/13/2025
Prosecution Insights