Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
DETAILED ACTION
The Action is responsive to the Amendments and Remarks filed on 3/9/2026. Claims 1-11 and 13-21 are pending claims. Claims 1, 10, and 20 are written in independent form. Claim 21 is new. Claim 12 has been cancelled.
Claim Objections
Claim 20 is objected to because of the following informalities:
Claim 20 appears to recite a typographical error by reciting “…the one or more randomly samples tokens…”. The intended language is understood to recite “the one or more randomly sampled [[samples]] tokens…” to be consistent with “one or more randomly sampled tokens” previously recited in the claims.
Appropriate correction is required.
Claim Interpretation
Claim 10 recites the phrases “to cause…to train” and “to perform” which is being understood as the intent to cause to train and the intent to perform, but is not actively performing any training or performing step/limitation. Examiner suggests to amend the claim limitations to recite all of the steps in a positive manner.
Claim 10 recites the limitation “provide the synthetic training set to a training engine to cause the training engine to train a machine learning model (MLM) to perform a domain-specific conversational task” which is being interpreted to have a scope of “provide the synthetic training set to a training engine”. However, for the purpose of compact prosecution, the limitation is being addressed herein as if all of the steps are recited in a positive manner.
Claim 19 recites the phrases “a system for performing” and “a system for generating or presenting” which are being understood as the intent to use the system for a particular purpose (performing, generating, or presenting), but is not actively performing any actual performing, generating, or presenting step/limitation. Examiner suggests to amend the claim limitations to recite all of the steps in a positive manner.
Claim 19 recites the limitation “wherein the system is comprised in at least one of…a system for performing simulation operations; a system for performing digital twin operations; a system for performing light transport simulation; a system for performing collaborative content creation for 3D assets; a system for performing deep learning operations;…a system for generating or presenting at least one of virtual reality content, mixed reality content, or augmented reality content;…a system for performing conversational AI operations; a system for generating synthetic data;” which is being interpreted as not further narrowing the scope. However, for the purpose of compact prosecution, the limitation is being addressed herein as if all of the steps are recited in a positive manner.
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claim 20 is rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention.
Regarding Claim 20, the claim recites an apparatus (“a processor performing”) without any components for performing the claimed function/steps because patentable weight is not being given to the preamble. The claim is therefore understood as being indefinite and is rejected under 35 U.S.C. 112(b) or pre-AIA 35 U.S.C. 112, second paragraph because it is unclear what components, if any, is performing the claimed function/steps.It is noted and restated herein that during the Examiner Interview on 1/20/2026 and included in the Examiner Interview Summary mailed on 1/23/2026, it was “highly recommended” to include in Claim 20 “language from the specification that a non-transitory computer readable medium contains instructions that when executed by one or more processors performs the claimed steps”. “Non-transitory computer-readable storage medium” that stores instructions and “one or more processors” executing the instructions stored on a computer-readable storage medium appears to have support in at least Paragraph [0110] of Applicant’s Specification.
The following is a quotation of 35 U.S.C. 112(d):
(d) REFERENCE IN DEPENDENT FORMS.—Subject to subsection (e), a claim in dependent form shall contain a reference to a claim previously set forth and then specify a further limitation of the subject matter claimed. A claim in dependent form shall be construed to incorporate by reference all the limitations of the claim to which it refers.
The following is a quotation of pre-AIA 35 U.S.C. 112, fourth paragraph:
Subject to the following paragraph [i.e., the fifth paragraph of pre-AIA 35 U.S.C. 112], a claim in dependent form shall contain a reference to a claim previously set forth and then specify a further limitation of the subject matter claimed. A claim in dependent form shall be construed to incorporate by reference all the limitations of the claim to which it refers.
Dependent Claim 3 is rejected under 35 U.S.C. 112(d) or pre-AIA 35 U.S.C. 112, 4th paragraph, as being of improper dependent form for failing to further limit the subject matter of the claim upon which it depends, or for failing to include all the limitations of the claim upon which it depends. Dependent Claim 3 recites “wherein the template query further comprises one or more placeholders, and wherein the modifying the template query comprises replacing the one or more placeholders with the one or more tokens” when their corresponding amended Independent Claim 1 already recites “modifying the selected template query by replacing one or more placeholder tokens in the selected template query with the one or more tokens”. Applicant may cancel the claim(s), amend the claim(s) to place the claim(s) in proper dependent form, rewrite the claim(s) in independent form, or present a sufficient showing that the dependent claim(s) complies with the statutory requirements.
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 1-11 and 13-21 are rejected under 35 U.S.C. 101 because the claimed invention is directed to non-patentable subject matter. The claimed invention is directed to one or more abstract ideas without significantly more. The judicial exception is not integrated into a practical application. The claims do not include additional elements that are sufficient to amount to significantly more than judicial exception. The eligibility analysis in support of these findings is provided below.
As per Claims 1, 10, and 20,
STEP 1:In accordance with Step 1 of the eligibility inquiry (as explained in MPEP 2106), the claimed method (claims 1-9 and 21), system (claims 10-11 and 13-19), and processor (claim 20) are directed to one of the eligible categories of subject matter and therefore satisfies Step 1.
STEP 2A Prong One:The independent claim 1 recites the following limitations directed to an abstract idea:
Performing one or more conversational tasks in a target domain using a machine learning model (MLM),
The limitation recites a mental process of observation, evaluation, judgement, and/or opinion capable of being performed by the human mind by making observation and evaluation of conversation in a target domain/topic and making a judgement and/or opinion of conversational responses (tasks) based on the observation and evaluation.
the MLM trained, at least, by:
generating a synthetic training set comprising a plurality of natural language (NL) queries,
The limitation recites a mental process of observation, evaluation, judgement, and/or opinion capable of being performed by the human mind by making a judgement and/or opinion to think of/generate a synthetic training set comprising a plurality of natural language queries.
at least one NL query of the plurality of NL queries generated, at least, by:
selecting a template query from a plurality of template queries,
The limitation recites a mental process of observation, evaluation, judgement, and/or opinion capable of being performed by the human mind by making an observation and evaluation of a plurality of template queries, and making a judgement and/or opinion to select one of the template queries based on the observation and evaluation.
sampling one or more tokens from a plurality of tokens corresponding to the target domain;
The limitation recites a mental process of observation, evaluation, judgement, and/or opinion capable of being performed by the human mind by making an observation and evaluation of a plurality of tokens corresponding to the target domain, and making a judgement and/or opinion of a sample set of one or more tokens based on the observation and evaluation.
modifying the selected template query by replacing one or more placeholder tokens in the selected template query with the one or more tokens to generate an NL query prompt comprising the one or more conversational entries with the one or more tokens inserted therein,
The limitation recites a mental process of observation, evaluation, judgement, and/or opinion capable of being performed by the human mind by observing and evaluating the selected template query comprising one or more conversational entries, one or more placeholder tokens in the selected template query, and the sampled one or more tokens, and making a judgement and/or opinion, based on the observation and evaluation, that modifies the selected template query into an NL query prompt comprising the one or more conversational entries with the one or more tokens inserted therein by replacing the placeholder tokens with the sampled one or more tokens.
Processing the NL query prompt to generate the at least one NL query of the plurality of NL queries,
The limitation recites a mental process of observation, evaluation, judgement, and/or opinion capable of being performed by the human mind by observing and evaluating an NL query prompt, and making a judgement and/or opinion to think of/generate a natural language query based on the observation and evaluation.
Updating one or more parameters of the MLM using the synthetic training set.
The limitation recites a mathematical concept of executing a mathematical function, such as a training, retraining, or any other updating algorithm, that updates parameters of a machine learning model using input, such as synthetic training data,
The independent claim 10 recites the following limitations directed to an abstract idea:
Generate a synthetic training set comprising a plurality of natural language (NL) queries, at least, by:
The limitation recites a mental process of observation, evaluation, judgement, and/or opinion capable of being performed by the human mind by making a judgement and/or opinion to think of/generate a synthetic training set comprising a plurality of natural language queries.
selecting a template query of a plurality of template queries;
The limitation recites a mental process of observation, evaluation, judgement, and/or opinion capable of being performed by the human mind by making an observation and evaluation of a plurality of template queries, and making a judgement and/or opinion to select one of the template queries based on the observation and evaluation.
sampling one or more tokens from a data store of domain-specific tokens;
The limitation recites a mental process of observation, evaluation, judgement, and/or opinion capable of being performed by the human mind by making an observation and evaluation of a plurality of tokens from a data store of domain-specific tokens, and making a judgement and/or opinion of a sample set of one or more tokens based on the observation and evaluation.
modifying the selected template query by replacing one or more placeholder tokens in the selected template query with the one or more tokens to generate an NL query prompt comprising the one or more conversational entries with the one or more tokens inserted therein,
The limitation recites a mental process of observation, evaluation, judgement, and/or opinion capable of being performed by the human mind by observing and evaluating the selected template query comprising one or more conversational entries, one or more placeholder tokens in the selected template query, and the sampled one or more tokens, and making a judgement and/or opinion, based on the observation and evaluation, that modifies the selected template query into an NL query prompt comprising the one or more conversational entries with the one or more tokens inserted therein by replacing the placeholder tokens with the sampled one or more tokens.
Processing the NL query prompt to generate a respective NL query of the plurality of NL queries,
The limitation recites a mental process of observation, evaluation, judgement, and/or opinion capable of being performed by the human mind by observing and evaluating an NL query prompt, and making a judgement and/or opinion to think of/generate a respective natural language query based on the observation and evaluation.
The independent claim 20 recites the following limitations directed to an abstract idea:
performing one or more natural language processing tasks in a target domain,
The limitation recites a mental process of observation, evaluation, judgement, and/or opinion capable of being performed by the human mind by making observation and evaluation of natural language (such as conversation) and a target domain, and making a judgement and/or opinion of the meaning or responses (tasks) based on the observation and evaluation.
the machine learning model trained, at least, using a synthetic dataset, and
The limitation recites a mathematical concept of executing a mathematical function, such as a training or retraining algorithm, that takes as input natural language queries previously generated using an LLM.
The one or more randomly sampled tokens associated with the target domain.
The limitation recites a mental process of observation, evaluation, judgement, and/or opinion capable of being performed by the human mind by observing and evaluating one or more randomly sampled tokens and the target domain, and based on the observation and evaluation, making a judgement and/or opinion that the one or more randomly sampled tokens are associated with the target domain.
STEP 2A Prong Two:Claim 1 recites that the method is performed using “a machine learning model (MLM)” and “a large language model (LLM)”, which is a high-level recitation of generic computer components and represents mere instructions to apply on a computer as in MPEP 2106.05(f), which does not provide integration into a practical application.
Claim 10 recites using “one or more processing units”, a machine learning model “MLM” and “a large language model (LLM)”, which is a high-level recitation of generic computer components and represents mere instructions to apply on a computer as in MPEP 2106.05(f), which does not provide integration into a practical application.
Claim 1 recites the following additional elements:
The selected template query comprising one or more conversational entries associated with the target domain;
The limitation recites an insignificant extra-solution activity as selecting a particular type of data being used to represent the selected template query as identified in MPEP 2106.05(g) and does not provide integration into a practical application.
The NL query prompt being a natural language text input to a large language model (LLM);
The limitation recites an insignificant extra-solution activity as selecting a particular type of data being used to represent the NL query prompt as identified in MPEP 2106.05(g) and does not provide integration into a practical application.
Providing the NL query prompt as input to the LLM;
The limitation recites an insignificant extra solution activity as sending/receiving of data (ie. Mere data gathering) such as ‘obtaining information’ as identified in MPEP 2106.05(g) and does not provide integration into a practical application.
The at least one NL query comprising the one or more additional conversational entries that (i) include at least some of the one or more sampled tokens and (ii) use, as a template, the one or more conversational entries;
The limitation recites an insignificant extra-solution activity as selecting a particular type of data being used to represent the at least one NL query, and subsequently the one or more additional conversational entries comprised within the at least one NL query, as identified in MPEP 2106.05(g) and does not provide integration into a practical application.
Claim 10 recites the following additional elements:
The selected template query comprising one or more conversational entries;
The limitation recites an insignificant extra-solution activity as selecting a particular type of data being used to represent the selected template query as identified in MPEP 2106.05(g) and does not provide integration into a practical application.
The NL query prompt being a natural language text input to a large language model (LLM);
The limitation recites an insignificant extra-solution activity as selecting a particular type of data being used to represent the NL query prompt as identified in MPEP 2106.05(g) and does not provide integration into a practical application.
Providing the NL query prompt as input to the LLM;
The limitation recites an insignificant extra solution activity as sending/receiving of data (ie. Mere data gathering) such as ‘obtaining information’ as identified in MPEP 2106.05(g) and does not provide integration into a practical application.
the respective NL query comprising the one or more additional conversational entries that (i) include at least some of the one or more sampled tokens and (ii) use, as a template, the one or more conversational entries; and
The limitation recites an insignificant extra-solution activity as selecting a particular type of data being used to represent the respective NL query, and subsequently the one or more additional conversational entries comprised within the respective NL query, as identified in MPEP 2106.05(g) and does not provide integration into a practical application.
provide the synthetic training set to a training engine to cause the training engine to train a machine learning model (MLM) to perform a domain-specific conversational task.
The limitation recites an insignificant extra solution activity as sending/receiving of data (ie. Mere data gathering) such as ‘obtaining information’ as identified in MPEP 2106.05(g) and does not provide integration into a practical application.
Claim 20 recites using “a processor”, “a machine learning model” and “a large language model (LLM)”, which is a high-level recitation of generic computer components and represents mere instructions to apply on a computer as in MPEP 2106.05(f), which does not provide integration into a practical application.
Claim 20 recites the following additional elements:
the synthetic dataset comprising natural language conversational entries generated by a large language model (LLM) responsive to NL prompts that comprise example NL conversational entries and one or more randomly sampled tokens inserted therein.
The limitation recites an insignificant extra-solution activity as selecting a particular type of data being used to represent the synthetic dataset and NL prompts, and subsequently the natural language conversational entries comprised within the synthetic dataset, as identified in MPEP 2106.05(g) and does not provide integration into a practical application.
Viewing the additional limitations together and the claim as a whole, nothing provides integration into a practical application.
STEP 2B:
The conclusions for the mere implementation using a computer are carried over and does not provide significantly more.
With respect to Claim 1 reciting “The selected template query comprising one or more conversational entries associated with the target domain;” identified as insignificant extra-solution activity above this is also WURC as court-identified see MPEP 2106.05(d)(II)(iv).
With respect to Claim 1 reciting “The NL query prompt being a natural language text input to a large language model (LLM);” identified as insignificant extra-solution activity above this is also WURC as court-identified see MPEP 2106.05(d)(II)(iv).
With respect to Claim 1 reciting “Providing the NL query prompt as input to the LLM;” identified as insignificant extra-solution activity above this is also WURC as court-identified see MPEP 2106.05(d)(II)(i).
With respect to Claim 1 reciting “The at least one NL query comprising the one or more additional conversational entries that (i) include at least some of the one or more sampled tokens and (ii) use, as a template, the one or more conversational entries;” identified as insignificant extra-solution activity above this is also WURC as court-identified see MPEP 2106.05(d)(II)(iv).
With respect to Claim 10 reciting “The selected template query comprising one or more conversational entries;” identified as insignificant extra-solution activity above this is also WURC as court-identified see MPEP 2106.05(d)(II)(iv).
With respect to Claim 10 reciting “Providing the NL query prompt as input to the LLM;” identified as insignificant extra-solution activity above this is also WURC as court-identified see MPEP 2106.05(d)(II)(i).
With respect to Claim 10 reciting “the respective NL query comprising the one or more additional conversational entries that (i) include at least some of the one or more sampled tokens and (ii) use, as a template, the one or more conversational entries;” identified as insignificant extra-solution activity above this is also WURC as court-identified see MPEP 2106.05(d)(II)(iv).
With respect to Claim 10 reciting “providing the generated plurality of NL queries to an MLM training engine to cause the MLM training engine to train a MLM to perform a domain-specific conversational task.” identified as insignificant extra-solution activity above this is also WURC as court-identified see MPEP 2106.05(d)(II)(i).
With respect to Claim 20 reciting “the synthetic dataset comprising natural language conversational entries generated by a large language model (LLM) responsive to prompts that comprise example NL conversational entries and one or more randomly sampled tokens.” identified as insignificant extra-solution activity above this is also WURC as court-identified see MPEP 2106.05(d)(II)(iv).
Looking at the claim as a whole does not change this conclusion and the claim is ineligible.
As per Dependent Claims 2-9, 11, 13-19, and 21,
STEP 1:In accordance with Step 1 of the eligibility inquiry (as explained in MPEP 2106), the claimed method (claims 1-9 and 21), system (claims 10-11 and 13-19), and processor (claim 20) are directed to one of the eligible categories of subject matter and therefore satisfies Step 1.
STEP 2A Prong One:The dependent claims 2-9, 11, 13-19, and 21 recite the following limitations directed to an abstract idea:
The limitation of Dependent Claim 3 includes the step(s) of:
wherein the modifying the template query comprises replacing the one or more placeholders with the one or more tokens.
The limitation recites a mental process of observation, evaluation, judgement, and/or opinion capable of being performed by the human mind by observing and evaluating the template query and the sampled one or more tokens, and making a judgement and/or opinion to replace placeholders with tokens, based on the observation and evaluation.
The limitation of Dependent Claims 5 and 14 includes the step(s) of:
wherein the sampling the one or more tokens comprises:
selecting a plurality of tokens based on correspondence of the plurality of tokens to the target domain; and
The limitation recites a mental process of observation, evaluation, judgement, and/or opinion capable of being performed by the human mind by observing and evaluating correspondence of the plurality of tokens to the target domain, and based on the observation and evaluation, making a judgement and/or opinion of selecting a plurality of tokens.
randomly sampling the one or more tokens from the selected plurality of tokens
The limitation recites a mental process of observation, evaluation, judgement, and/or opinion capable of being performed by the human mind by making an observation the plurality of tokens, and based on the observation, making a judgement and/or opinion of randomly sampling the plurality of tokens.
The limitation of Dependent Claims 6 and 15 includes the step(s) of:
wherein at least one template query of the plurality of template queries is associated with a selection weight, and
The limitation recites a mental process of observation, evaluation, judgement, and/or opinion capable of being performed by the human mind by observing and evaluating the template query and making a judgement and/or opinion to associate a selection weight based on the observation and evaluation.
wherein the template query is selected with a probability determined using the selection weight.
The limitation recites a mental process of observation, evaluation, judgement, and/or opinion capable of being performed by the human mind by observing a template query and a probability determined using the selection weight, and making a judgement and/or opinion to select the template query based on the observation and evaluation.
The limitation of Dependent Claims 7 and 16 includes the step(s) of:
augmenting the NL query prompt with the obtained condition to generate an augmented query prompt;
The limitation recites a mental process of observation, evaluation, judgement, and/or opinion capable of being performed by the human mind by observing and evaluating the obtained condition and the query prompt, and making a judgement and/or opinion that modifies/augments the query prompt, based on the observation and evaluation.
applying the augmented NL query prompt to the LLM to generate the at least one NL query.
The limitation recites a mental process of observation, evaluation, judgement, and/or opinion capable of being performed by the human mind by observing and evaluating a query prompt that has been augmented/modified, and making a judgement and/or opinion to think of/generate a natural language query based on the observation and evaluation.
The limitation of Dependent Claims 8 and 17 includes the step(s) of:
the obtaining the condition comprises:
selecting a subset of at least one of the plurality of tokens; and
The limitation recites a mental process of observation, evaluation, judgement, and/or opinion capable of being performed by the human mind by observing and evaluating a plurality of tokens and making a judgement and/or opinion to select a subset of the plurality of tokens based on the observation and evaluation.
combining the subset with a condition template associated with the sub-task,
The limitation recites a mental process of observation, evaluation, judgement, and/or opinion capable of being performed by the human mind by observing and evaluating a subset of tokens and a condition template, and making a judgement and/or opinion to combine them based on the observation and evaluation.
wherein the training the MLM includes using the subset.
The limitation recites a mental process of observation, evaluation, judgement, and/or opinion capable of being performed by the human mind by at least observing and evaluating the subset of tokens during training of the MLM.
The limitation of Dependent Claims 9 and 18 includes the step(s) of:
associating a quality label with one or more NL queries, the quality label being indicative of a quality of a respective NL query of the one or more NL queries;
The limitation recites a mental process of observation, evaluation, judgement, and/or opinion capable of being performed by the human mind by observing and evaluating one or more NL queries, and making a judgement and/or opinion of a quality label, indicative of a quality of the observed and evaluated NL queries, to be associated with the one or more NL queries.
creating a new template query using one or more template queries of the plurality of template queries, the new template query comprising one or more virtual tokens;
The limitation recites a mental process of observation, evaluation, judgement, and/or opinion capable of being performed by the human mind by observing and evaluating one or more template queries and based on the observation and evaluation, making an judgement and/or opinion of a new template query and virtual tokens included in the new template query.
using another MLM to learn the one or more virtual tokens of the new template query using the one or more NL queries and associated quality labels; and
The limitation recites a mental process of observation, evaluation, judgement, and/or opinion capable of being performed by the human mind by observing and evaluating the new template query, one or more NL queries, and associated quality labels, and making a judgement and/or opinion to learn/remember the virtual tokens of the new template query based on the observation and evaluation.
The limitation of Dependent Claim 21 includes the step(s) of:
wherein the LLM completes the dynamic input/output pairing by generating an output value based on the one or more example input/output pairings as few-shot examples.
The limitation recites a mathematical concept of executing a mathematical function in the form of an LLM that takes as input at least “one or more example input/output pairings as few-shot examples” and the NL query prompt including one or more example input/output pairings in a static portion and a dynamic input/output pairing with the one or more sampled tokens inserted and an empty output value, and outputs “the dynamic input/output pairing by generating an output value”.
STEP 2A Prong Two:The claim(s) recite the following additional elements:
The limitation of Dependent Claims 2 and 11 includes the step(s) of:
wherein the template query comprises:
an example query corresponding to the target domain, and
an NL response to the example query.
The limitation recites an insignificant extra-solution activity as selecting a particular type of data being used to represent a query template as identified in MPEP 2106.05(g) and does not provide integration into a practical application.
The limitation of Dependent Claim 3 includes the step(s) of:
wherein the template query further comprises one or more placeholders,
The limitation recites an insignificant extra-solution activity as selecting a particular type of data being used to represent a query template as identified in MPEP 2106.05(g) and does not provide integration into a practical application.
The limitation of Dependent Claims 4 and 13 includes the step(s) of:
wherein the example query corresponds to an example conversation between a customer and a service provider in at least one of:
a financial services domain,
a food services domain,
a health services domain, or
a retail services domain.
The limitation recites an insignificant extra-solution activity as selecting a particular type/domain of data being used to represent an example query as identified in MPEP 2106.05(g) and does not provide integration into a practical application.
The limitation of Dependent Claims 7 and 16 includes the step(s) of:
obtaining a condition associated with a sub-task of the conversational task;
The limitation recites an insignificant extra solution activity as retrieval of data (ie. Mere data gathering) such as ‘obtaining information’ as identified in MPEP 2106.05(g) and does not provide integration into a practical application.
The limitation of Dependent Claims 8 and 17 includes the step(s) of:
wherein the one or more tokens includes a plurality of tokens,
The limitation recites an insignificant extra-solution activity as selecting a particular type of data (more than one/a plurality of tokens) being used to represent the one or more tokens as identified in MPEP 2106.05(g) and does not provide integration into a practical application.
The limitation of Dependent Claims 9 and 18 includes the step(s) of:
adding the new template query comprising the one or more learned virtual tokens to the plurality of template queries.
The limitation recites an insignificant extra solution activity as retrieval of data (ie. Mere data gathering) such as ‘obtaining information’ as identified in MPEP 2106.05(g) and does not provide integration into a practical application.
The limitation of Dependent Claim 19 includes the step(s) of:
wherein the system is comprised in at least one of:
an in-vehicle infotainment system for an autonomous or semi-autonomous machine;
a system for performing simulation operations;
a system for performing digital twin operations;
a system for performing light transport simulation;
a system for performing collaborative content creation for 3D assets;
a system for performing deep learning operations;
a system for generating or presenting at least one of virtual reality content, mixed reality content, or augmented reality content;
a system for performing conversational AI operations; or
a system for generating synthetic data;
The limitation recites an insignificant extra-solution activity as selecting a particular type of data/system being used to represent the system as identified in MPEP 2106.05(g) and does not provide integration into a practical application.
wherein the system is comprised in at least one of:
a system implemented using an edge device;
a system implemented using a robot;
a system incorporating one or more virtual machines (VMs);
a system implemented at least partially in a data center; or
a system implemented at least partially using cloud computing resources.
The limitation recites a high-level recitation of generic computer components and represents mere instructions to apply on a computer as in MPEP 2106.05(f), which does not provide integration into a practical application.
The limitation of Dependent Claim 21 includes the step(s) of:
wherein the NL query prompt includes one or more example input/output pairings in a static portion and a dynamic input/output pairing with the one or more sampled tokens inserted and an empty output value,
The limitation recites an insignificant extra-solution activity as selecting a particular type of data (one or more example input/output pairings in a static portion and a dynamic input/output pairing with the one or more sampled tokens inserted and an empty output value) being used to represent the NL query prompt as identified in MPEP 2106.05(g) and does not provide integration into a practical application.
Viewing the additional limitations together and the claim as a whole, nothing provides integration into a practical application.
STEP 2B:
The conclusions for the mere implementation using a computer are carried over and does not provide significantly more.
With respect to Claims 2 and 11 reciting “wherein the template query comprises: an example query corresponding to the target domain, and an NL response to the example query;” identified as insignificant extra-solution activity above this is also WURC when claimed in a merely generic manner as court-identified see MPEP 2106.05(d)(II)(iv).
With respect to Claim 3 reciting “wherein the template query further comprises one or more placeholders,” identified as insignificant extra-solution activity above this is also WURC when claimed in a merely generic manner as court-identified see MPEP 2106.05(d)(II)(iv).
With respect to Claims 4 and 13 reciting “wherein the example query corresponds to an example conversation between a customer and a service provider in at least one of: a financial services domain, a food services domain, a health services domain, or a retail services domain.” identified as insignificant extra-solution activity above this is also WURC when claimed in a merely generic manner as court-identified see MPEP 2106.05(d)(II)(iv).
With respect to Claims 7 and 16 reciting “obtaining a condition associated with a sub-task of the conversational task;” identified as insignificant extra-solution activity above this is also WURC as court-identified see MPEP 2106.05(d)(II)(i).
With respect to Claims 8 and 17 reciting “wherein the one or more tokens includes a plurality of tokens,” identified as insignificant extra-solution activity above this is also WURC when claimed in a merely generic manner as court-identified see MPEP 2106.05(d)(II)(iv).
With respect to Claims 9 and 18 reciting “adding the new template query comprising the one or more learned virtual tokens to the plurality of template queries.” identified as insignificant extra-solution activity above this is also WURC as court-identified see MPEP 2106.05(d)(II)(i).
With respect to Claim 19 reciting “wherein the system is comprised in at least one of: an in-vehicle infotainment system for an autonomous or semi-autonomous machine; a system for performing simulation operations; a system for performing digital twin operations; a system for performing light transport simulation; a system for performing collaborative content creation for 3D assets; a system for performing deep learning operations; a system for generating or presenting at least one of virtual reality content, mixed reality content, or augmented reality content; a system for performing conversational AI operations; or a system for generating synthetic data;” identified as insignificant extra-solution activity above this is also WURC when claimed in a merely generic manner as court-identified see MPEP 2106.05(d)(II)(iv).
With respect to Claim 21 reciting “wherein the NL query prompt includes one or more example input/output pairings in a static portion and a dynamic input/output pairing with the one or more sampled tokens inserted and an empty output value,” identified as insignificant extra-solution activity above this is also WURC when claimed in a merely generic manner as court-identified see MPEP 2106.05(d)(II)(iv).
Looking at the claim as a whole does not change this conclusion and the claim is ineligible.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for 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 20 is rejected under 35 U.S.C. 103 as being unpatentable over Arthur et al. (U.S. Pre-Grant Publication No. 2023/0186161, hereinafter referred to as Arthur) and further in view of Dai et al. (International Publication No. WO2024/064249, hereinafter referred to as Dai).
Regarding Claim 20:
Arthur teaches:
a processor performing:
Arthur teaches “Some embodiments include a system that includes one or more data processors; and one or more non-transitory computer-readable media storing instructions which, when executed by the one or more processors, cause the one or more processors to perform part or all of the operations and/or methods disclosed herein.” (Para. [0027]).
one or more natural language processing tasks in a target domain using a machine learning model,
Arthur teaches “A bot (also referred to as a skill, chatbot, chatterbot, or talkbot) is a computer program that can perform conversations with end users. The bot can generally respond to natural-language messages (e.g., questions or comments) through a messaging application that uses natural-language messages. Enterprises may use one or more bots to communicate with end users through a messaging application.” (Para. [0055])
Arthur further teaches “In certain examples, the skill bot uses a predictive model that is trained using the training data and allows the skill bot to discern what users say (or in some cases, are trying to say). DABP 102 provides various different training techniques that can be used by a skill bot designer to train a skill bot, including various machine-learning based training techniques, rules-based training techniques, and/or combinations thereof.” (Para. [0094]).Arthur also teaches the tasks being in a target domain by teaching, respectively, a financial services domain, a food services domain, a health services domain, and a retail services domain:
(financial services domain) “end users interact with the bot through conversation interactions” and “End users also interact with the bot through other types of interactions, such as transactional interactions (e.g., with a banking bot that is at least trained to transfer money from one account to another), informational interactions (e.g., with a human resources bot that is at least trained check the remaining vacation hours the user has), and/or retail interactions (e.g., with a retail bot that is at least trained for discussing returning purchased goods or seeking technical support).” (Para. [0057])
(food services domain) “an owner of a restaurant (e.g., a pizza shop) may use DABP 102 to create and deploy a digital assistant that enables customers of the restaurant to order food (e.g., order pizza)” (Para. [0061]).
(health services domain) “Databases power information systems across multiple industries including retail (e.g., orders, cancellations, refunds), supply chain (e.g., raw materials, stocks, vendors), healthcare (e.g., medical records), and finance (e.g., financial business metrics) to name a few” (Para. [0047]).
(retail services domain) “end users interact with the bot through conversation interactions” and “End users also interact with the bot through other types of interactions, such as transactional interactions (e.g., with a banking bot that is at least trained to transfer money from one account to another), informational interactions (e.g., with a human resources bot that is at least trained check the remaining vacation hours the user has), and/or retail interactions (e.g., with a retail bot that is at least trained for discussing returning purchased goods or seeking technical support).” (Para. [0057])
the machine learning model trained, at least, using a synthetic dataset
Arthur teaches “the synthetic training data can be combined with original training data in order to train a machine learning model” (Abstract).
The NL prompts comprise:
one or more randomly sampled tokens inserted therein, the one or more randomly sampled tokens associated with the target domain.
Arthur teaches randomly sampled tokens associated with a target domain by teaching “in order to sample components for a respective analyzed template of the templates 4210, a database of the one or more databases 4216 can be randomly selected and its components can be sampled and used to lexicalize the respective analyzed template” (Para. [0150]) where the template is associated with a particular task or target domain.
Arthur explicitly teaches all of the elements of the claimed invention as recited above except:
The synthetic dataset comprising natural language conversational entries generated by a large language model (LLM) responsive to NL prompts
The NL prompts comprise example NL conversational entries
However, in the related field of endeavor of generating a synthetic training dataset, Dai teaches:
The synthetic dataset comprising natural language conversational entries generated by a large language model (LLM) responsive to NL prompts, and
Dai teaches “The PROMPTAGATOR may be configured to transform a few examples into many more examples by prompting an LLM (e.g., LLM 120) to generate more data (e.g., synthetic training dataset 130)” (Para. [56]) and “Data augmentation via synthetic query generation generally involves question generators” (Para. 42]) thereby teaching providing a prompt for a NL query/question to an LLM that generates NL queries/questions based on the input.
The NL prompts comprise example NL conversational entries
Dai teaches “The PROMPTAGATOR may be configured to transform a few examples into many more examples by prompting an LLM (e.g., LLM 120) to generate more data (e.g., synthetic training dataset 130)” (Para. [56]) and “Data augmentation via synthetic query generation generally involves question generators” (Para. 42]) thereby teaching providing a prompt for a NL query/question to an LLM that generates NL queries/questions based on the input.Dai further teaches “A few-shot LLM query generator described herein can produce good queries without any fine-tuning the model. In fact, synthetically generated data can be strong enough to reduce and/or eliminate the use of annotated query-document pairs from traditional high-resource datasets such as Natural Questions.” (Para. [43]) where a query Q can be “e.g., short keyword search queries, questions, arguments, etc.” (Para. [051]).
Thus, it would have been obvious to one of ordinary skill in the art, having the teachings of Dai and Arthur at the time that the claimed invention was effectively filed, to have modified the trained chatbot system, as taught by Arthur, with the amplification of example questions for generating a synthetic training dataset, as taught by Dai.
One would have been motivated to make such combination because Dai teaches “examples, amplified by prompt-based LLM query generation, simplifies the complexity of training neural retrievers for new tasks and leads to significant performance gains” (para. [189]) and it would have been obvious to a person having ordinary skill in the art that “simplifying the complexity of training” and achieving “significant performance gains” would be a desirable general goal when training a machine learning model.
Claim(s) 1-11, 13-19, and 21 are rejected under 35 U.S.C. 103 as being unpatentable over Mishra et al. (U.S. Pre-Grant Publication No. 2021/0342552, hereinafter referred to as Mishra) and further in view of Arthur et al. (U.S. Pre-Grant Publication No. 2023/0186161, hereinafter referred to as Arthur) and Dai et al. (International Publication No. WO2024/064249, hereinafter referred to as Dai).
Regarding Claim 1:
Mishra teaches a method comprising:
Generating a synthetic training set comprising a plurality of natural language (NL) queries,
Mishra further teaches “once processing of the sequence of template tags is complete, the resulting natural language content (e.g., sentence, clause, phrase, etc.) is generated from the selected word at operation 470” (Para. [0105]).
Mishra further teaches that the natural language content can be a natural language question/query by teaching example generated output as questions (Fig. 9) and the natural language content is understood as synthetic because it was generated and did not previously exist.
at least one NL query of the plurality of NL queries generated, at least, by:
selecting a template query from a plurality of template queries,
Mishra teaches “client systems 114 enable users to submit sets of keywords and templates (and optionally part-of-speech (POS) tags for the keywords) to server systems 110 for generation of natural language content (e.g., sentences, clauses, phrases, etc.)” (Para.[0037]) where “templates in a natural language (even though finite) may be considerably large in number (e.g., millions, etc.).” (Para. [0030]). Therefore, Mishra teaches selecting a template from a plurality of templates.Mishra further teaches that the template can be a template query by teaching example template sentences as questions or requests (Fig. 9).
The selected template query comprising one or more conversational entries associated with the target domain;
Mishra teaches the intent to transform input text into user-specified stylized text for a particular domain (Para. [0005]) where “Present invention embodiments employ a control specification, in the form of templates, that uses an input-style agnostic approach for which a domain expert is not required” (Para. [0030]). Therefore, Mishra teaches the selected template query comprising entries associated with a particular domain that would have otherwise required a domain expert.Mishra further teaches the templates comprising entries that are “conversational entries” by teaching example template sentences comprising conversational sentences/entries such as “who is bob’s friend?” and “that is not a good book” (Fig. 9). It is noted that a conversational entry is not defined in the specification and is therefore being given its broadest reasonable interpretation as an entry relating to natural language words/topics used in conversation.
sampling one or more tokens from a plurality of tokens corresponding to the target domain;
Mishra teaches “keywords are associated with language tags, while the template includes a series of language tags indicating an arrangement for the generated natural language content” (Para. [0007]) where “the set of keywords may include any quantity of any tokens (e.g., words, acronyms, abbreviations, etc.) of any length from any natural language, and may be arranged in any desired order” (Para. [0146]) thereby teaching taking a sample of one or more of the “any quantity of any tokens”.
Mishra further teaches “The various vectors for keywords, keyword tags, and template tags (e.g., embeddings, transformed vectors, encoded template vectors, attended context vectors, etc.) may include any quantity of elements or features of any values to indicate a context or meaning for the corresponding item (e.g., keyword, tag, set of keywords, etc.).” (Para. [0148]) thereby teaching the tokens corresponding to a context/domain.
modifying the selected template query by replacing one or more placeholder tokens in the selected template query with the one or more tokens to generate an NL query prompt
Mishra teaches “the tag embedding module further generates tag embeddings or vector representations for the POS tags of the template at operation 425, while template encoder 345 produces encoded vectors for the template tags at operation 430 representing context of the template tags” (Para. [0099]) thereby teaching modifying the selected template to generate a NL query prompt.Mishra further teaches “one or more vocabulary words are selected for the template tag of the current time step at operation 460” (Para. [0103]) thereby teaching the template having a placeholder for the “one or more vocabulary words” selected and replacing the placeholder tokens with the “one or more vocabulary words”.
It is noted that a natural language (NL) query prompt is being understood as a prompt for generating a natural language query, which is consistent with later claimed steps, and does not require that the query prompt is in a natural language format. Therefore, Mishra teaches the query prompt as a prompt for a natural language query by teaching that the prompt will be used result in “output summarization text… generated using a deep learning-based natural language generation (NLG) approach. This approach enables output text to be semantically related to the input data, thereby conveying the desired meaning. Further, the output text is transformed according to the template, and knowledge of adhering to the template style may be obtained from easily implementable custom or conventional natural language processing (NLP) systems (e.g., deterministic or machine learning based classifiers), regressors, and/or metrics.” (Para. [0032]) and “once processing of the sequence of template tags is complete, the resulting natural language content (e.g., sentence, clause, phrase, etc.) is generated from the selected word at operation 470” (Para. [0105]).Providing the NL query prompt as input to the model for processing;
Mishra teaches “output summarization text is generated using a deep learning-based natural language generation (NLG) approach…the output text is transformed according to the template, and knowledge of adhering to the template style may be obtained from easily implementable custom or conventional natural language processing (NLP) systems (e.g., deterministic or machine learning based classifiers), regressors, and/or metrics.” (Para. [0032]) and “once processing of the sequence of template tags is complete, the resulting natural language content (e.g., sentence, clause, phrase, etc.) is generated from the selected word at operation 470” (Para. [0105]).Therefore, Mishra teaches providing the NL query prompt to the natural language generation model which will then process the prompt/input and output natural language content.
processing the NL query prompt to generate the at least one NL query of the plurality of NL queries,
Mishra teaches “output summarization text is generated using a deep learning-based natural language generation (NLG) approach. This approach enables output text to be semantically related to the input data, thereby conveying the desired meaning. Further, the output text is transformed according to the template, and knowledge of adhering to the template style may be obtained from easily implementable custom or conventional natural language processing (NLP) systems (e.g., deterministic or machine learning based classifiers), regressors, and/or metrics.” (Para. [0032]).
Mishra further teaches “once processing of the sequence of template tags is complete, the resulting natural language content (e.g., sentence, clause, phrase, etc.) is generated from the selected word at operation 470” (Para. [0105]).
Mishra explicitly teaches all of the elements of the claimed invention as recited above except:
Performing one or more conversational tasks in a target domain using a machine learning model (MLM),
Providing the NL query prompt as input to the LLM;
The NL query prompt comprising the one or more conversational entries with the one or more tokens inserted therein, the NL query being a natural langue text input to a large language model (LLM);
processing, using the LLM, the query prompt
The at least one NL query comprising the one or more additional conversational entries that (i) include at least some of the one or more sampled tokens and (ii) use, as a template, the one or more conversational entries; and
Updating one or more parameters of the MLM using the synthetic training set.
However, in the related field of endeavor of synthesizing synthetic natural language data, Arthur teaches:
Performing one or more conversational tasks in a target domain using a machine learning model (MLM),
Arthur teaches “A bot (also referred to as a skill, chatbot, chatterbot, or talkbot) is a computer program that can perform conversations with end users. The bot can generally respond to natural-language messages (e.g., questions or comments) through a messaging application that uses natural-language messages. Enterprises may use one or more bots to communicate with end users through a messaging application.” (Para. [0055])
Arthur further teaches “In certain examples, the skill bot uses a predictive model that is trained using the training data and allows the skill bot to discern what users say (or in some cases, are trying to say). DABP 102 provides various different training techniques that can be used by a skill bot designer to train a skill bot, including various machine-learning based training techniques, rules-based training techniques, and/or combinations thereof.” (Para. [0094]).
Updating one or more parameters of the MLM using the synthetic training set.
Arthur teaches “the synthetic training data can be combined with original training data in order to train a machine learning model” (Abstract).
Thus, it would have been obvious to one of ordinary skill in the art, having the teachings of Arthur and Mishra at the time that the claimed invention was effectively filed, to have modified the systems and methods for natural language text generation, as taught by Mishra, with the trained chatbot system, as taught by Arthur,
One would have been motivated to make such combination because while Mishra teaches generating natural language text in one context of menu screens (Para. [0037] - “the client systems may present a graphical user(e.g., GUI, etc.) or other interface (e.g., command line prompts, menu screens, etc.) to solicit information from users”) and Arthur teaches using generated synthetic natural language text to train digital assistants or chatbots to perform specific skills, such as “ordering food” (Para. [0071]) and it would have been obvious to a person having ordinary skill in the art that creating a digital assistant or a chatbot using at least the generated natural language text taught by Mishra would expand the use-case of the menu screen client systems taught by Mishra for an improved user/client experience.
Arthur and Mishra explicitly teach all of the elements of the claimed invention as recited above except:
Providing the NL query prompt as input to the LLM;
The NL query prompt comprising the one or more conversational entries with the one or more tokens inserted therein, the NL query being a natural langue text input to a large language model (LLM);
processing, using the LLM, the query prompt;
The at least one NL query comprising the one or more additional conversational entries that (i) include at least some of the one or more sampled tokens and (ii) use, as a template, the one or more conversational entries;
However, in the related field of endeavor of generating a synthetic training dataset, Dai teaches:
Providing the NL query prompt as input to a large language model (LLM);
Dai teaches “The PROMPTAGATOR may be configured to transform a few examples into many more examples by prompting an LLM (e.g., LLM 120) to generate more data (e.g., synthetic training dataset 130)” (Para. [56]) and “Data augmentation via synthetic query generation generally involves question generators” (Para. 42]) thereby teaching providing a prompt for a NL query/question to an LLM that generates NL queries/questions based on the input.
The NL query prompt comprising the one or more conversational entries with the one or more tokens inserted therein, the NL query being a natural langue text input to a large language model (LLM);
Dai teaches conversational entries by teaching “such a prompt may be run on documents from T).sub.T 235 to generate a large set of synthetic (q, d) examples 220, amplifying the information from a few examples (e.g., task-specific few-shot examples 225) into a large synthetic dataset (e.g., synthetic data 220)” (Para. [63]) and “The few-shot examples, amplified by prompt-based LLM query generation, simplifies the complexity of training neural retrievers for new tasks and leads to significant performance gains” (para. [189]) and Mishra teaches inserting tokens into a query (“one or more vocabulary words are selected for the template tag of the current time step at operation 460” - Para. [0103])Dai further teaches “natural questions-QGen” and “T5 Qgen model finetuned on NQ” (Para. [88]) as well as “Natural Questions (NQ)” (Para. [90]) thereby teaching the NL query being natural language text input to an LLM.
processing, using the LLM, the query prompt;
Dai teaches “such a prompt may be run on documents from T).sub.T 235 to generate a large set of synthetic (q, d) examples 220, amplifying the information from a few examples (e.g., task-specific few-shot examples 225) into a large synthetic dataset (e.g., synthetic data 220)” (Para. [63]) and “The few-shot examples, amplified by prompt-based LLM query generation, simplifies the complexity of training neural retrievers for new tasks and leads to significant performance gains” (para. [189]) thereby teaching processing the query prompt using an LLM.
The at least one NL query comprising the one or more additional conversational entries that (i) include at least some of the one or more sampled tokens and (ii) use, as a template, the one or more conversational entries;
Dai teaches “PROMPTAGATOR can use a few prompts that enable few-shot retrieval, resulting in a significant improvement with just two to eight examples for each task, by inputting these new examples in the prompt” (Para. [40]).Mishra teaches “keywords are associated with language tags, while the template includes a series of language tags indicating an arrangement for the generated natural language content” (Para. [0007]) where “the set of keywords may include any quantity of any tokens (e.g., words, acronyms, abbreviations, etc.) of any length from any natural language, and may be arranged in any desired order” (Para. [0146]) and example template sentences comprising conversational sentences/entries such as “who is bob’s friend?” and “that is not a good book” including the tokens and using conversational entries as a template (Fig. 9).
Therefore, Dai in combination with Mishra teaches the additional conversational entries (Dai – Para. [40]) in the NL query including sampled tokens and using, as a template, conversational entries (Mishra – Paras. [0007], [0146], & Fig. 9).
Thus, it would have been obvious to one of ordinary skill in the art, having the teachings of Dai, Arthur, and Mishra at the time that the claimed invention was effectively filed, to have modified the trained chatbot system, as taught by Arthur, and the systems and methods for natural language text generation, as taught by Mishra, with the amplification of example questions for generating a synthetic training dataset, as taught by Dai.
One would have been motivated to make such combination because Dai teaches “examples, amplified by prompt-based LLM query generation, simplifies the complexity of training neural retrievers for new tasks and leads to significant performance gains” (para. [189]) and it would have been obvious to a person having ordinary skill in the art that “simplifying the complexity of training” and achieving “significant performance gains” would be a desirable general goal when training a machine learning model.
Regarding Claim 2:
Dai, Arthur and Mishra further teach:
wherein the template query comprises:
an example query corresponding to the target domain, and
Mishra teaches an Example Template Sentence, Template POS, and Generated Output as part of a template query (Fig. 9).
an NL response to the example query.
Mishra teaches an Example Template Sentence, Template POS, and Generated Output as part of a template query (Fig. 9).
Regarding Claim 3:
Dai, Arthur and Mishra further teach:
wherein the template query further comprises one or more placeholders, and
Mishra teaches “one or more vocabulary words are selected for the template tag of the current time step at operation 460” (Para. [0103]) thereby teaching the template having a placeholder for the “one or more vocabulary words” selected.
wherein the modifying the template query comprises replacing the one or more placeholders with the one or more tokens.
Mishra teaches “one or more vocabulary words are selected for the template tag of the current time step at operation 460” (Para. [0103]) thereby teaching the template having a placeholder and replacing it with the “one or more vocabulary words” selected.
Regarding Claim 4:
Dai, Arthur and Mishra further teach:
wherein the example query corresponds to an example conversation between a customer and a service provider in at least one of:
a financial services domain,
Arthur teaches “end users interact with the bot through conversation interactions” and “End users also interact with the bot through other types of interactions, such as transactional interactions (e.g., with a banking bot that is at least trained to transfer money from one account to another), informational interactions (e.g., with a human resources bot that is at least trained check the remaining vacation hours the user has), and/or retail interactions (e.g., with a retail bot that is at least trained for discussing returning purchased goods or seeking technical support).” (Para. [0057])
a food services domain,
Arthur teaches “an owner of a restaurant (e.g., a pizza shop) may use DABP 102 to create and deploy a digital assistant that enables customers of the restaurant to order food (e.g., order pizza)” (Para. [0061]).
a health services domain, or
Arthur teaches “Databases power information systems across multiple industries including retail (e.g., orders, cancellations, refunds), supply chain (e.g., raw materials, stocks, vendors), healthcare (e.g., medical records), and finance (e.g., financial business metrics) to name a few” (Para. [0047]).
a retail services domain.
Arthur teaches “end users interact with the bot through conversation interactions” and “End users also interact with the bot through other types of interactions, such as transactional interactions (e.g., with a banking bot that is at least trained to transfer money from one account to another), informational interactions (e.g., with a human resources bot that is at least trained check the remaining vacation hours the user has), and/or retail interactions (e.g., with a retail bot that is at least trained for discussing returning purchased goods or seeking technical support).” (Para. [0057])
Regarding Claim 5:
Dai, Arthur and Mishra further teach:
wherein the sampling the one or more tokens comprises:
selecting a plurality of tokens based on correspondence of the plurality of tokens to the target domain; and
Arthur teaches “constrained sampling stage 4212, components are sampled from the database 4216 based on a database analysis 4218 of the database 4216 and the analyzed templates 4210. In some instances, the database 4216 is one or more relational databases with each database having components (e.g., tables, columns, and values).” (Para. [0149]).
randomly sampling the one or more tokes from the selected plurality of tokens.
Arthur teaches performing random sampling by teaching “in order to sample components for a respective analyzed template of the templates 4210, a database of the one or more databases 4216 can be randomly selected and its components can be sampled and used to lexicalize the respective analyzed template” (Para. [0150]).
Regarding Claim 6:
Dai, Arthur and Mishra further teach:
wherein at least one template query of the plurality of template queries is associated with a selection weight, and
Mishra teaches “the probability (e.g., likelihood of matching a template tag to a keyword tag) may be expressed in any manner (e.g., value ranges, percentages, etc.) indicating a probability in a range of 0% to 100%. The similarity may be measured using any desired similarity or distance measure (e.g., cosine similarity, Euclidean distance, etc.). The attention weights may be any suitable values (e.g., indicating a probability in the range of 0% to 100%, etc.) to adjust contributions of the keywords to the attended context.” (Para. [0149]).
wherein the template query is selected with a probability determined using the selection weight.
Mishra teaches a “probability for each language tag of the template indicating a likelihood of that language tag of the template matching one of the associated language tags of the keywords” (Para. [0053]) where “the probability (e.g., likelihood of matching a template tag to a keyword tag) may be expressed in any manner (e.g., value ranges, percentages, etc.) indicating a probability in a range of 0% to 100%. The similarity may be measured using any desired similarity or distance measure (e.g., cosine similarity, Euclidean distance, etc.). The attention weights may be any suitable values (e.g., indicating a probability in the range of 0% to 100%, etc.) to adjust contributions of the keywords to the attended context.” (Para. [0149]).
Regarding Claim 7:
Dai, Arthur and Mishra further teach:
obtaining a condition associated with a sub-task of the conversational task;
Mishra teaches “present invention embodiments handle templates of various sentence forms, such as declarative, interrogative, exclamatory, and negation” (Para. [0136]) and “Once processing of the sequence of template tags is complete, the set of words indicated by the probability distributions of GRUs 359 forms the resulting natural language content (e.g., sentence, clause, phrase, etc.). The form of the words in the resulting content may be adjusted or modified to align with the forms indicated by the template tags (e.g., plural, possessive, verb conjugation, etc.). The word vocabulary may include additional function words beyond the keywords and/or template tags” (Para. [0094]) thereby teaching receiving a condition associated with a sub-task and adjusting or modifying based on obtained conditions.
augmenting the NL query prompt with the obtained condition to generate an augmented NL query prompt; and
Mishra teaches “present invention embodiments handle templates of various sentence forms, such as declarative, interrogative, exclamatory, and negation” (Para. [0136]) and “Once processing of the sequence of template tags is complete, the set of words indicated by the probability distributions of GRUs 359 forms the resulting natural language content (e.g., sentence, clause, phrase, etc.). The form of the words in the resulting content may be adjusted or modified to align with the forms indicated by the template tags (e.g., plural, possessive, verb conjugation, etc.). The word vocabulary may include additional function words beyond the keywords and/or template tags” (Para. [0094]) thereby teaching receiving a condition associated with a sub-task and adjusting or modifying based on obtained conditions.
applying the NL augmented query prompt to the LLM to generate the at least one NL query.
Mishra teaches “present invention embodiments handle templates of various sentence forms, such as declarative, interrogative, exclamatory, and negation” (Para. [0136]) and “Once processing of the sequence of template tags is complete, the set of words indicated by the probability distributions of GRUs 359 forms the resulting natural language content (e.g., sentence, clause, phrase, etc.). The form of the words in the resulting content may be adjusted or modified to align with the forms indicated by the template tags (e.g., plural, possessive, verb conjugation, etc.). The word vocabulary may include additional function words beyond the keywords and/or template tags” (Para. [0094]) thereby teaching receiving a condition associated with a sub-task and adjusting or modifying based on obtained conditions to generate the output natural language content.
Regarding Claim 8:
Dai, Arthur and Mishra further teach:
wherein the one or more tokens includes a plurality of tokens, and
Mishra teaches “keywords are associated with language tags, while the template includes a series of language tags indicating an arrangement for the generated natural language content” (Para. [0007]) where “the set of keywords may include any quantity of any tokens (e.g., words, acronyms, abbreviations, etc.) of any length from any natural language, and may be arranged in any desired order” (Para. [0146]).
the obtaining the condition comprises:
selecting a subset of at least one of the plurality of tokens; and
Mishra teaches “keywords are associated with language tags, while the template includes a series of language tags indicating an arrangement for the generated natural language content” (Para. [0007]) where “the set of keywords may include any quantity of any tokens (e.g., words, acronyms, abbreviations, etc.) of any length from any natural language, and may be arranged in any desired order” (Para. [0146])
combining the subset with a condition template associated with the sub-task,
Mishra teaches “The probability for a corresponding language tag of the template indicates the contribution for the context of the keywords for generating a word for the corresponding language tag of the template, and a complement of the probability indicates the contribution for the context of the template for generating the word for the corresponding language tag of the template. This enables a decoder to shift focus between keyword representations and generic language (e.g., POS, etc.) representations based on the language (e.g., POS, etc.) representations in the template.” (Para. [0010]) thereby teaching combining the subset of tokens with the template for the particular context.
wherein the training the MLM includes using the subset.
Mishra teaches “output summarization text is generated using a deep learning-based natural language generation (NLG) approach. This approach enables output text to be semantically related to the input data, thereby conveying the desired meaning. Further, the output text is transformed according to the template, and knowledge of adhering to the template style may be obtained from easily implementable custom or conventional natural language processing (NLP) systems (e.g., deterministic or machine learning based classifiers), regressors, and/or metrics.” (Para. [0032]).
Arthur teaches “the synthetic training data can be combined with original training data in order to train a machine learning model”.Therfore, Mishra in combination with Arthur teaches using generated synthetic natural language data to train a machine learning model.
Regarding Claim 9:
Dai, Arthur and Mishra further teach:
associating a quality label with one or more NL queries, the quality label being indicative of a quality of a respective NL query of the one or more NL queries;
Arthur teaches “at validation stage 4224, each lexicalized training example of lexicalized training examples 4222 can be validated and lexicalized training examples that are valid can be included in the lexicalized training data 4114 (i.e., the synthetic training data) and lexicalized training examples that are not valid can be discarded (i.e., the discarded training examples 4226)” (Para. [0153]) thereby teaching associating a quality/validation label with the training data indicative of a quality/validation of the training data. Arthur further teaches the training data comprising synthetic natural language questions/queries (Para. [0183]).
creating a new template query using one or more template queries of the plurality of template queries, the new template query comprising one or more virtual tokens;
Mishra teaches “Templates can be created from sentences of a large unlabeled text corpus. In other words, any sentence can be used to create a template.” (Para. [0007]).
Mishra further teaches “keywords are associated with language tags, while the template includes a series of language tags indicating an arrangement for the generated natural language content” (Para. [0007]) where “the set of keywords may include any quantity of any tokens (e.g., words, acronyms, abbreviations, etc.) of any length from any natural language, and may be arranged in any desired order” (Para. [0146]) thereby teaching newly created templates having virtual tokens.
using another MLM to learn the one or more virtual tokens of the new template query using the one or more NL queries and associated quality labels; and
Arthur teaches “Environment 100 comprises a digital assistant builder platform (DABP) 102 that enables users 104 of DABP 102 to create and deploy digital assistants or chatbot systems. DABP 102 can be used to create one or more digital assistants (or DAs) or chatbot systems. For example, as shown in FIG. 1, users 104 representing a particular enterprise can use DABP 102 to create and deploy a digital assistant 106 for users of the particular enterprise. For example, DABP 102 can be used by a bank to create one or more digital assistants for use by the bank's customers. The same DABP 102 platform can be used by multiple enterprises to create digital assistants. As another example, an owner of a restaurant (e.g., a pizza shop) may use DABP 102 to create and deploy a digital assistant that enables customers of the restaurant to order food (e.g., order pizza).” (Para. [0061]) thereby teaching training different chatbots or digital assistants on different NL queries generated from template queries for the different enterprise purposes.
adding the new template query comprising the one or more learned virtual tokens to the plurality of template queries.
Mishra teaches “The system may employ any number of any conventional or other databases, data stores or storage structures (e.g., files, databases, data structures, data or other repositories, etc.) to store any information (e.g., training data, vocabulary, machine learning or other parameters, dictionaries, etc.). The database system may be implemented by any number of any conventional or other databases, data stores or storage structures (e.g., files, databases, data structures, data or other repositories, etc.) to store information.” (Para. [0143]).
Regarding Claim 10:
Some of the limitations herein are similar to some or all of the limitations of Claim 1.
Dai, Arthur and Mishra further teach:
one or more processing units (Mishra – Para. [0158]).
Regarding Claim 11:
All of the limitations herein are similar to some or all of the limitations of Claim 2.
Regarding Claim 13:
All of the limitations herein are similar to some or all of the limitations of Claim 4.
Regarding Claim 14:
All of the limitations herein are similar to some or all of the limitations of Claim 5.
Regarding Claim 15:
All of the limitations herein are similar to some or all of the limitations of Claim 6.
Regarding Claim 16:
All of the limitations herein are similar to some or all of the limitations of Claim 7.
Regarding Claim 17:
All of the limitations herein are similar to some or all of the limitations of Claim 8.
Regarding Claim 18:
All of the limitations herein are similar to some or all of the limitations of Claim 9.
Regarding Claim 19:
Dai, Arthur and Mishra further teach:
wherein the system is comprised in at least one of:
an in-vehicle infotainment system for an autonomous or semi-autonomous machine;
Arthur teaches “A digital assistant can be embodied or implemented in various physical systems or devices, such as in a computer, a mobile phone, a watch, an appliance, a vehicle, and the like.” (Para. [0062]).
a system for performing simulation operations;
Arthur teaches “This analysis may be performed by one or more processors, possibly processing the data in parallel, performing simulations using the data, and the like” (Para. [0249]).
a system for performing digital twin operations;
Arthur teaches “This analysis may be performed by one or more processors, possibly processing the data in parallel, performing simulations using the data, and the like” (Para. [0249]).
a system for performing light transport simulation;
Arthur teaches “This analysis may be performed by one or more processors, possibly processing the data in parallel, performing simulations using the data, and the like” (Para. [0249]).
a system for performing collaborative content creation for 3D assets;
a system for performing deep learning operations;
Mishra teaches “using a deep learning-based natural language generation (NLG) approach” (Para. [0032]).
a system implemented using an edge device;
Mishra teaches the network may comprise…edge servers” (Para [0155]).
a system for generating or presenting at least one of virtual reality content, mixed reality content, or augmented reality content;
a system implemented using a robot;
Arthur teaches “A bot (also referred to as a skill, chatbot, chatterbot, or talkbot) is a computer program that can perform conversations with end users. The bot can generally respond to natural-language messages (e.g., questions or comments) through a messaging application that uses natural-language messages. Enterprises may use one or more bots to communicate with end users through a messaging application.” (Para. [0055]).
a system for performing conversational AI operations;
Arthur teaches “The intents and their associated example utterances are used as training data to train the skill bot. Various different training techniques may be used. As a result of this training, a predictive model is generated that is configured to take an utterance as input and output an intent inferred for the utterance by the predictive model. In some instances, input utterances are provided to an intent analysis engine, which is configured to use the trained model to predict or infer an intent for the input utterance. The skill bot may then take one or more actions based upon the inferred intent.” (Para. [0091])
a system for generating synthetic data;
Arthur teaches “Techniques are disclosed herein for synthesizing synthetic training data” (Abstract).
a system incorporating one or more virtual machines (VMs);
Arthur teaches “server 612 may include one or more virtual machines” (Para. [0233]).
a system implemented at least partially in a data center; or
Mishra teaches examples of computer systems, environments, and/or configurations that may be suitable for use with computer system 212 include, but are not limited to…distributed cloud computing environments” (Para. [0042]) thereby teaching a data center for the cloud computing environments.
a system implemented at least partially using cloud computing resources.
Mishra teaches examples of computer systems, environments, and/or configurations that may be suitable for use with computer system 212 include, but are not limited to…distributed cloud computing environments” (Para. [0042]).
Regarding Claim 21:
Dai, Arthur and Mishra further teach:
the NL query prompt includes one or more example input/output pairings in a static portion and a dynamic input/output pairing with the one or more sampled tokens inserted and an empty output value,
Dai further teaches using “few-shot prompting with GPT-3 to generate synthetic data for training” (Para. [105]) which few-shot prompting is understood as using a prompt that necessarily includes pairs of inputs and desired outputs in the prompt to define instructions/examples for a desired output for an input with an empty/unknown output.
wherein the LLM completes the dynamic input/output pairings by generating an output value based on the one or more example input/output pairings as few-shot examples.
Dai further teaches completing the dynamic input/output pairings by teaching using “few-shot prompting with GPT-3 to generate synthetic data for training” (Para. [105]).
Response to Amendment
Applicant’s Amendments, filed on 3/9/2026, are acknowledged and accepted.
In light of the amendments filed on 3/9/2026, the 112(f) interpretation of Claim 20 has been withdrawn. However, Claim 20 is still rejected under 35 U.S.C. 112(b) or pre-AIA 35 U.S.C. 112, second paragraph as being is indefinite because it is unclear what structure, if any, is related to the claimed parts, as is further explained in the rejection above.
Response to Arguments
On page 11 of the Remarks filed on 3/9/2026, Applicant states, with respect to the 101 rejection, that “claims 1-20 do not recite concepts that explicitly fall into the abstract idea exception categories of (a) mathematical concepts, (b) certain methods of organizing human activity, or (c) mental processes. Therefore for at least this reason, the eligibility analysis should end at the first prong of Revised Step 2A.”.Applicant’s statement is not convincing and the amended claims hare are fully addressed in the rejection above.
On page 11 of the Remarks filed on 3/9/2026, Applicant argues, with respect to the 101 rejection, that “the practical application of the claims lies in generating synthetic training data for training conversational machine learning models by using a specific technical structure that is provided to a large language model (LLM) to generate linguistically diverse conversational entries, as described in paragraphs [0017]-[0018] and [0038]- [0041] of the Specification”.Applicant’s argument is not convincing because while the claims recite “generating a synthetic training set comprising a plurality of natural language (NL) queries” where only “at least one NL query of the plurality of NL queries [is] generated” by the rest of the limitations and “updating one or more parameters of the MLM using the synthetic training set”, there is no indication that the generated “at least one NL query” in the synthetic training data is utilized for any part of the updating because the synthetic training set comprises “a plurality of natural language (NL) queries”.
On pages 11-12 of the Remarks filed on 3/9/2026, Applicant argues, with respect to the 101 rejection, that “New claim 21 describes additional details of the specific technical structure (e.g., "the NL query prompt includes one or more example input/output pairings in a static portion and a dynamic input/output pairing with the one or more sampled tokens inserted and an empty output value, wherein the LLM completes the dynamic input/output pairing by generating an output value based on the one or more example input/output pairings as few-shot examples"). The techniques described in the Specification enable the generation of robust and varied synthetic training datasets that address the technological problem identified in paragraph [0016] of the Specification, where "generative grammars or template forms are used to create synthetic training data. However, this often leads to a lack of linguistic diversity due to the restrictive nature of templates and to accidental inclusion of ungrammatical or implausible examples." The claimed invention overcomes this technological limitation by using an LLM (e.g., to complete a dynamic input/output pairing based on few-shot examples provided in the prompt), which generates outputs that "more closely simulates a likely human's response" (paragraph [0041]) including "conversational colloquialisms such as asking for 'the twelve-piece' instead of 'the twelve-piece chicken bucket"' (paragraph [0041]). The resulting synthetic training data enables the conversational MLM to "appropriately respond to a large variety of user interactions" (paragraph [0020])”.Applicant’s argument is not convincing because newly added Claim 21 is merely understood as reciting an additional element and “wherein the NL query prompt includes…” and a mathematical concept of executing a mathematical function in the form of an LLM, as is further addressed in the rejection above.
On pages 12-13 of the Remarks filed on 3/9/2026, Applicant argues, with respect to the 101 rejection, that “the claims add significantly more than the judicial exception under Step 2B. The amended claims recite a specific ordered combination of steps that is not well-understood, routine, or conventional.”Applicant’s argument has been fully considered but was not found to be convincing in overcoming the rejection presented above based on the amended claims.
On page 13 of the Remarks filed on 3/9/2026, Applicant argues, with respect to the 101 rejection, that “the specific technical structure of the NL query prompt recited in new claim 21-comprising a static portion with example input/output pairings and a dynamic portion with domain-specific sampled tokens and an empty output value-and the use of an LLM to complete the dynamic portion via few-shot learning represents a specific technical approach to synthetic training data generation that is neither routine nor conventional.”Applicant’s argument has been fully considered but was not found to be convincing in overcoming the rejection presented above based on the amended claims.
On page 13 of the Remarks filed on 3/9/2026, Applicant argues, with respect to the 101 rejection, that “As described in paragraph [0016] of the Specification, prior approaches using “generative grammars or template forms" to create synthetic training data "often leads to a lack of linguistic diversity due to the restrictive nature of templates and to accidental inclusion of ungrammatical or implausible examples." The claimed approach overcomes these limitations by leveraging the LLM's few-shot learning capability to generate linguistically diverse conversational entries that enable the trained MLM to "appropriately respond to a large variety of user interactions" (paragraph [0020]). This represents a concrete technical improvement over prior template-based synthetic data generation methods.”Applicant’s argument is not convincing because Claim 21 merely recites the contents of the prompt and that the is LLM executed to generate an output value using the LLM’s few-shot learning capability as part of the execution, but does not relate how the output value is utilized in any meaningful way that might integrate the claims into a practical application.
On page 14 of the Remarks filed on 3/9/2026, Applicant argues, with respect to the 103 rejection, that “Arthur is directed to data manufacturing frameworks for synthesizing synthetic training data to facilitate training a natural language to logical form model. (Arthur, Abstract.) Arthur teaches generating synthetic training data by lexicalizing templates i.e., replacing non-terminal symbols in delexicalized utterances with sampled database components such as table names, column names, and values. (Arthur, paragraphs [0146], [0148].) However, Arthur's lexicalized utterances are the output of a data manufacturing process for training a natural language to SQL model-they are not NL prompts provided to an LLM to generate additional conversational entries.” and “The Office Action asserts that Dai teaches the features not taught by Arthur, namely, the synthetic dataset comprising natural language conversational entries generated by a large language model (LLM) responsive to NL prompts comprising example NL conversational entries. Applicant respectfully disagrees.” because “Dai does not remedy the shortcomings of Arthur with respect to claim 20. Dai is directed to systems and methods for prompt-based query generation for diverse retrieval. (Dai, Abstract.) Dai teaches using few-shot query-document examples as prompts to an LLM to generate synthetic query-document pairs for training a document retrieval model. (Dai, paragraphs [42]-[43], [56].) However, Dai's prompts are few-shot examples provided to the LLM as-is.”Applicant’s argument is not convincing because Arthur was never relied upon alone as teaching NL prompts provided to an LLM to generate additional conversational entries. Further, it is Arthur in combination with Dai that teaches the features Applicant appears to be arguing, Arthur teaching constructing the NL prompts with randomly sampled tokens inserted therein (so that they are not “as-is” as Applicant is arguing) and Dai teaching synthetic data comprising natural language conversational entries generated by an LLM responsive to NL prompts.
On page 14 of the Remarks filed on 3/9/2026, Applicant argues, with respect to the 103 rejection, that “Additionally, as recited in new claim 21, Dai does not teach NL prompts that comprise "one or more example input/output pairings in a static portion and a dynamic input/output pairing with the one or more randomly sampled tokens inserted and an empty output value," nor does Dai teach an LLM that "completes the dynamic input/output pairing by generating an output value based on the one or more example input/output pairings as few-shot examples." Thus, Dai does not cure the deficiencies of Arthur.”Upon further consideration of Dai based on the newly claimed features in Claim 21, Dai was found to teach the features which are addressed fully in the rejection above.
On pages 15-16 of the Remarks filed on 3/9/2026, Applicant argues, with respect to the 103 rejection, that Claim 1 has been amended to “directly address the Examiner's concerns. Claim 1, as amended, now specifies (1) that modification occurs "by replacing one or more placeholder tokens in the selected template query with the one or more sampled tokens,"” and that the “amendments specify how the tokens modify the template query and define the prompt as natural language text which excludes Mishra's vector embedding approach.” where “Mishra does not teach creating a natural language text prompt by replacing placeholder tokens with domain-specific tokens, where the resulting prompt comprises conversational entries with the sampled tokens inserted therein.”Upon further review of the amended language and the previously cited prior art, it appears that Applicant used language from dependent claims 3 and 12 and incorporated these into the Independent Claims. The cited portions with respect to these dependent claims were still found to teach the amended language as is fully addressed in the rejection above.
On page 16 of the Remarks filed on 3/9/2026, Applicant argues, with respect to the 103 rejection, that “The amendments to claim 1 also go beyond dependent claim 3, which recites "wherein the template query further comprises one or more placeholders, and wherein the modifying the template query comprises replacing the one or more placeholders with the one or more tokens." Claim 1, as amended, additionally specifies (1) the structure of the resulting prompt (comprising conversational entries with tokens inserted), and (2) that the prompt is a natural language text input to the LLM.”.Applicant’s argument is moot because Claim 3 is understood as not furthering the scope from amended language in claim 1, as is addressed in more detail in the 112(d) rejection above.
On page 17 of the Remarks filed on 3/9/2026, Applicant argues, with respect to the 103 rejection, that Neither Dai nor Arthur teaches the features of Claim 21 reciting “that the prompt includes example input/output pairings in a static portion and a dynamic input/output pairing with an empty output value that the LLM completes based on few-shot examples. These additional elements further distinguish over Mishra's embedding-level approach.”.Upon further consideration of Dai based on the newly claimed features in Claim 21, Dai was found to teach the features which are addressed fully in the rejection above.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Reza et al. (U.S. Pre-Grant Publication No. 2023/0237277) teaches dynamically developing a contextual set of prompts based on relevant aspects extracted from s set of training data. One technique includes obtaining training data comprising text examples and associated labels, extracting aspects from the training data, generating prompting templates based on the training data and the extracted aspects, concatenating each of the text examples with the respective generated prompting template to create prompting functions, training a machine learning language model on the prompting functions to predict a solution for a task, where the training is formulated as a masked language modeling problem with blanks of the prompting templates being set as text labels and expected output for the task being set as specified solution labels, and the training learns or updates model parameters of the machine learning language model for performing the task. The machine learning language model is provided with the learned or updated model parameters.
Cohen et al. (U.S. Pre-Grant Publication No. 2022/0261631) teaches provisioning of pipelines for efficient training, retraining, configuring, deploying, and using machine learning models for inference in user-specific platforms.The reference further teaches “Presently, configuring a MLM for applications in a user-specific domain may require significant development efforts. Developers of a MLM may need to design an architecture of the model (e.g., a number of layers and a topology of node connections, in case of a neural network MLM), train the MLM(s) on relevant domain-specific training data, and so on” (Para. [0016]).The reference further teaches “Retraining may be performed using retraining data tailored for a user-specific domain of use. In some embodiments, the retraining data may be provided by the user. For example, a user may provide retraining data to enhance natural language processing capabilities of one of the pre-trained MLMs to improve recognition of speech that may be encountered in an investment brokerage environment or a securities trading environment.” (Para. [0022]).
Bhardwaj et al. (U.S. Pre-Grant Publication No. 2023/0342559) teaches a soft prompt tuning technique referred to as the Vector quantized Input-contextualized Prompt (VIP). The VIP techniques has two integral properties i) instead of learning a fixed set of prompt tokens irrespective of the input, it generates a contextualized version of the soft prompts, conditional on the input text ii) it further passes the input-contextualized prompt tokens through a quantization network, inspired by Vector Quantized Transformers. The quantization network uses nearest neighbor search over a learnable codebook to train a discrete latent variable model over the prompt-space, thus generating quantized version of contextual prompt tokens. These quantized contextual prompt tokens are finally fed into the frozen language model along with the original input text.
Rodrigo Cavalin et al. (U.S. Pre-Grant Publication No. 2023/0092274) teaches a topic for building a new intent on which to train a chatbot can be received. A database of chatbot training data can be searched for a candidate intent having meta-knowledge similar to the received topic. Utterances associated with the candidate intent can be extracted. The received topic and the extracted utterances can be input to a trained machine learning model. The trained machine learning model generates example utterances for the new intent. The new intent with the generated example utterances can be used as training data for training the chatbot.
Earle et al. (U.S. Pre-Grant Publication No. 2024/0111960) teaches generating a synthetic chat between a customer module and an agent module, wherein: the customer module receives a first prompt and determines a first chat response, and the agent module receives a second prompt and determines a second chat response; generating, by a summarizer module, a summary of the synthetic chat; scoring, by a scorer module, the synthetic chat by comparing the summary of the synthetic chat to the first prompt and the second prompt; adjusting, based on the score, a parameter associated with the synthetic chat.The reference further teaches “the prompt generator 106 augments the prompts associated with each LLM of each interlocutor every iteration of the synthetic data generation.” (Para. [0024]).
Lester et al. (U.S. Pre-Grant Publication No. 2023/0325725) teaches natural language processing can leverage trained prompts to condition a large pre-trained machine-learned model to generate an output for a specific task. For example, a subset of parameters may be trained for the particular task to then be input with a set of input data into the pre-trained machine-learned model to generate the task-specific output. During the training of the prompt, the parameters of the pre-trained machine-learned model can be frozen, which can reduce the computational resources used during training while still leveraging the previously learned data from the pre-trained machine-learned model.The reference further teaches “prompt tuning in order to generate prompts associated with particular tasks to enable the use of pre-trained machine-learned models without retraining the large pre-trained machine-learned model.” (Para. [0001]).
Liu et al. (U.S. Patent NO. 11,948,563) teaches receiving a user request from a client system associated with a user, determining that the user request corresponds to a first suspended task, retrieving a first dialog state of the first suspended task from a dialog history associated with the user, generating a summary of the first suspended task based on the first dialog state using a natural-language generating (NLG) module, and sending instructions to the client system for providing the summary of the first suspended task to the user.
McVeigh et al. (U.S. Pre-Grant Publication No. 2026/0023787) teaches a conversational assistant that may be configured for use with one or more applications. The disclosed systems and methods may be configured as an outer loop/inner loop architecture incorporating one or more large language models (LLMs) and/or RAG functionality to support dialog-based interactions.
Hubli et al. (U.S. Pre-Grant Publication No. 2026/0119844) teaches automatic generation of log message templates from log messages can be performed and enhanced using trained discriminator model. Expression processor can separate main content portion from structured information items of log message. For log message, sequence of encoded tokens can be generated. Some encoded tokens can be replaced with masked tokens to generate masked sequence. Generator model can be trained to predict encoded tokens that were replaced with masked tokens in masked sequence. Modified sequence, comprising some encoded tokens and some predicted tokens replacing other encoded tokens, can be generated. Discriminator model can be trained to predict whether token in modified sequence is original encoded token or replaced token. For subsequent log message, trained discriminator model can infer whether token associated with log message is a dynamic token, and, if so, dynamic token can be replaced with defined character during generation of log message template.
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to ROBERT F MAY whose telephone number is (571)272-3195. The examiner can normally be reached Monday-Friday 9:30am to 6pm.
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, Boris Gorney can be reached on 571-270-5626. 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.
/ROBERT F MAY/Examiner, Art Unit 2154 5/14/2026
/BORIS GORNEY/Supervisory Patent Examiner, Art Unit 2154
Prosecution Insights