DETAILED ACTION
Remarks
Claims 1-12 and 15-22 have been examined and rejected. This Office Action is responsive to the amendment filed on 12/17/2025, which has been entered in the above identified application.
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 .
Claims 1, 2, 4, 6-11, 15 and 17-26 are presented for examination.
Response to Amendment
Applicant’s amendment filed on 12/17/2025 has been entered. Claims 1, 2, 4, 7, 8, 10 and 19-22 are amended. Claims 3, 5, 12 and 16 are cancelled. Claims 23-26 are added. Claims 1, 2, 4, 6-11, 15 and 17-26 are pending in the application.
Claim Rejections - 35 USC § 112
The following is a quotation of the first paragraph of 35 U.S.C. 112(a):
(a) IN GENERAL.—The specification shall contain a written description of the invention, and of the manner and process of making and using it, in such full, clear, concise, and exact terms as to enable any person skilled in the art to which it pertains, or with which it is most nearly connected, to make and use the same, and shall set forth the best mode contemplated by the inventor or joint inventor of carrying out the invention.
The following is a quotation of the first paragraph of pre-AIA 35 U.S.C. 112:
The specification shall contain a written description of the invention, and of the manner and process of making and using it, in such full, clear, concise, and exact terms as to enable any person skilled in the art to which it pertains, or with which it is most nearly connected, to make and use the same, and shall set forth the best mode contemplated by the inventor of carrying out his invention.
Claims 1, 2, 4, 6-11, 15 and 17-26 are rejected under 35 U.S.C. 112(a) or 35 U.S.C. 112 (pre-AIA ), first paragraph, as failing to comply with the written description requirement. The claim(s) contains subject matter which was not described in the Specification in such a way as to reasonably convey to one skilled in the relevant art that the inventor or a joint inventor, or for applications subject to pre-AIA 35 U.S.C. 112, the inventor(s), at the time the application was filed, had possession of the claimed invention.
With respect to claims 1 [line 2], claim 19 [line 2] and claim 20 [line 2], the limitation “receiving content of a completed virtual agent dialog that has occurred between at least two communication parties to resolve a task” are not disclosed in the Applicant’s Specification and are considered new matter. The Applicant’s Specification do not contain any description of a completed virtual agent dialog that has occurred between two communication parties. At best, Applicant’s Specification indicates that content of a dialog between at least two communication parties to resolve a task is received [par. 0012, 0038]; the content of the dialog is a collection of utterances in text format [par. 0038]. For the purposes of examination, Examiner will interpret the limitations as “receiving content of a virtual agent dialog between at least two communication parties to resolve a task.”
With respect to claims 1 [line 7], claim 19 [line 7] and claim 20 [line 7], the limitation “a name of the respective eligible step and a natural language description of the respective eligible step” are not disclosed in the Applicant’s Specification and are considered new matter. The Applicant’s Specification do not contain any description of a natural language description of the respective eligible step. At best, Applicant’s Specification indicates that Flow: followed by workflow step descriptions joined by a comma [par. 0022]; step prediction is possible because of the natural language training of the machine language models described [par. 0025]; text-to-text model 204 can also select out-of-domain workflow steps due to the natural language training of text-to-text model 204 [par. 0028]. For the purposes of examination, Examiner will interpret the limitations as “a name of the respective eligible step and a description of the respective eligible step due to the natural language training of text-to-text model.”
With respect to claims 2, 4, 6-11, 15, 17, 18, and 21-26, they are also rejected based on their virtual dependency of claims 1, 19 and 20.
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, 2, 4, 6-11, 15 and 17-26 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Independent claims
Step 1
Claim 1 is drawn to a method, claim 19 is drawn to a system comprising one or more processors and a memory coupled to processors, and claim 20 is drawn to a computer program product embodied in a non-transitory computer-readable medium comprising instructions when executed cause the processors to perform method of claim 1. Therefore, each of these claim groups falls under one of four categories of statutory subject matter (process/method, machines/product/apparatus, manufactures, and composition of matter).
Step 2A – Prong 1
Claims 1, 19 and 20 are directed to a judicially recognized exception of an abstract idea without significantly more.
Claims 1, 19 and 20 recite a method of receiving content of a completed virtual agent dialog that has occurred between at least two communication parties to resolve a task that under its broadest reasonable interpretation enumerates a mental concept. A human can mentally perform, with the physical aid such as pen and paper, to receive a virtual agent dialog between communication devices. Therefore, the step of receiving content of a completed virtual agent dialog is nothing more than a mental concept (MPEP 2106.04(a)(2)(III)).
Claims 1, 19 and 20 recite further a method of receiving a workflow domain specification associated with at least a portion of eligible steps of a workflow relating to the task, wherein the workflow domain specification comprises, for each respective eligible step of at least the portion of eligible steps, a name of the respective eligible step and a natural language description of the respective eligible step that under its broadest reasonable interpretation enumerates a mental concept. A human can mentally perform, with the physical aid such as pen and paper, to receive a specification with eligible steps of workflow. Therefore, the step of receiving a specification is nothing more than a mental concept (MPEP 2106.04(a)(2)(III)).
Claims 1, 19 and 20 recite further a method of determining machine learning model input data based on (i) the content of the completed virtual agent dialog and (ii) the workflow domain specification that under its broadest reasonable interpretation enumerates a mental concept. A human can mentally perform, with the physical aid such as pen and paper, to determine input data. Therefore, the step of determining machine learning input data based on content of the virtual agent is nothing more than a mental concept (MPEP 2106.04(a)(2)(III)).
Claims 1, 19 and 20 recite further a method of generating a representation of the sequence of workflow steps that under its broadest reasonable interpretation enumerates a mental concept. A human can mentally perform, with the physical aid such as pen and paper, to represent a sequence of workflow steps. Therefore, the step of generating representation of the sequence of workflow steps is nothing more than a mental concept (MPEP 2106.04(a)(2)(III)).
Step 2A – Prong 2
Claims 1, 19 and 20 recite further processing the machine learning model input data using a trained machine learning model executing on one or more hardware processors to generate a sequence of workflow steps that summarizes the completed virtual agent dialog, wherein the trained machine learning model is conditioned by the workflow domain specification to select one or more workflow steps of the sequence of workflow steps from the eligible steps of the workflow domain specification, and wherein, during training, the trained machine learning model had no exposure to at least one workflow step of the one or more workflow steps those fails to integrate the abstract idea into a practical application. The steps of processing the machine learning (ML) input data, using trained ML model, executing on hardware processors, trained ML model is conditioned and had no exposure are some forms of insignificant input and output solution activities, where processing ML model input; using trained ML model; executing on one or more hardware processors; trained ML model is conditioned by the workflow domain specification; and trained ML model had no exposure to the workflow step are necessary for all uses of the judicial exception. These additional elements do not integrate the abstract idea into a practical application because they do not impose any meaningful limits on practicing the abstract idea (MPEP 2106.05(g)).
Step 2B
The additional elements in step 2A-Prong 2 those are forms of insignificant extra-solution activities, do not amount to significantly more than an abstract idea because the court decision have determined that these additional elements of processing ML model input; using trained ML model; executing on one or more hardware processors; trained ML model is conditioned by the workflow domain specification; and trained ML model had no exposure to the workflow step to be well-understood, routine, and conventional when claimed in a merely generic manner (MPEP 2106.05(d)(II)).
As such, claims 1, 19 and 20 are not patent eligible.
Dependent claims
Claims 2, 4, 6-11, 15, 17, 18 and 21-26 merely narrow the previously recited abstract idea limitations. For the reasons described above with respect to claim 1, this judicial exception is not meaningfully integrated into a practical application, or significantly more than the abstract idea. The claims disclose similar limitations described for the independent claims above and do not provide anything more than the mental and mathematical processes that are practically capable of being performed in the human mind with the assistance of pen and paper. Therefore, claims 2, 4, 6-11, 15, 17, 18 and 21-26 also recite abstract ideas that do not integrate into a practical application or amount to significantly more than the judicial exception, and are rejected under U.S.C. 101.
Step 1
Claims 2, 4, 6-11, 15, 17, 18 and 21-26 are drawn to a method. Therefore, this claim group falls under one of four categories of statutory subject matter (process/method, machines/product/apparatus, manufactures, and composition of matter).
Step 2A – Prong 1
Dependent claim 10 recites further the mental process by wherein determining the machine learning model input data includes combining (i) the content of the completed virtual agent dialog and (ii) the workflow domain specification according to a specific textual format that is based on one or more features of the ML project (MPEP 2106.04(a)(2)(III)).
Dependent claim 25 recites further the mental process by wherein determining the specified list of specification options comprises: partitioning a dataset of training dialogs into a plurality of batches, wherein each respective batch of the plurality of batches includes semantically similar training dialogs; and processing each respective batch of the plurality of batches using the second machine learning model to predict a corresponding set of steps for each respective batch those are based on one or more features of the ML project (MPEP 2106.04(a)(2)(III)).
Dependent claim 26 recites further the mental process by wherein determining the specified list of specification options further comprises: determining a combined set of steps by aggregating the corresponding set of steps of each respective batch; and removing duplicates and semantically equivalent steps from the combined set of steps those are based on one or more features of the ML project (MPEP 2106.04(a)(2)(III)).
Step 2A – Prong 2
Dependent claim 2 recites further the insignificant extra solution activities by wherein the content of the completed virtual agent dialog comprises a plurality of natural language utterances arranged in a sequential time order associated with when the utterances occurred. This additional element does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea (MPEP 2106.05(g)).
Dependent claim 4 recites further the insignificant extra solution activities by wherein the at least two communication parties include (i) at least one communication party that is a virtual agent and (ii) zero or more communication parties that are human. This additional element does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea (MPEP 2106.05(g)).
Dependent claim 6 recites further the insignificant extra solution activities by wherein the task includes a customer support task. This additional element does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea (MPEP 2106.05(g)).
Dependent claim 7 recites further the insignificant extra solution activities by wherein the workflow domain specification has been selected from a specified list of workflow domain specification options. This additional element does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea (MPEP 2106.05(g)).
Dependent claim 8 recites further the insignificant extra solution activities by wherein the specified list of workflow domain specification options has been determined using a second machine learning model that has been trained to automatically predict a workflow steps domain based on an input dialog. This additional element does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea (MPEP 2106.05(g)).
Dependent claim 9 recites further the insignificant extra solution activities by wherein each step of the at least the portion of eligible steps is semantically related to the task. This additional element does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea (MPEP 2106.05(g)).
Dependent claim 11 recites further the insignificant extra solution activities by wherein the trained machine learning model is a text-to-text pre- trained language model. This additional element does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea (MPEP 2106.05(g)).
Dependent claim 15 recites further the insignificant extra solution activities by wherein the trained machine learning model has been further trained on an additional dataset to perform a summarization task. This additional element does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea (MPEP 2106.05(g)).
Dependent claim 17 recites further the insignificant extra solution activities by wherein the trained machine learning model has been further trained to perform a workflow discovery task. This additional element does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea (MPEP 2106.05(g)).
Dependent claim 18 recites further the insignificant extra solution activities by wherein the sequence of workflow steps comprises a plurality of textual descriptions of actions taken in sequential order to resolve the task. This additional element does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea (MPEP 2106.05(g)).
Dependent claim 21 recites further the insignificant extra solution activities by wherein at least one workflow step of the sequence of workflow steps includes respective parameters that relate to the task. This additional element does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea (MPEP 2106.05(g)).
Dependent claim 22 recites further the insignificant extra solution activities by wherein conditioning the trained machine learning model by the workflow domain specification constrains the trained machine learning model to select the one or more workflow steps from the eligible steps of the workflow domain specification. This additional element does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea (MPEP 2106.05(g)).
Dependent claim 23 recites further the insignificant extra solution activities by wherein the sequence of workflow steps comprises an invented step generated by the trained machine learning model, wherein the invented step is not present in the workflow domain specification, and wherein the trained machine learning model is configured to generate a new step name for the invented step. This additional element does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea (MPEP 2106.05(g)).
Dependent claim 24 recites further the insignificant extra solution activities by wherein during training, the trained machine learning model had no exposure to the invented step. This additional element does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea (MPEP 2106.05(g)).
As such, dependent claims 2, 4, 6-11, 15, 17, 18 and 21-26 are not patent eligible.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claims 1, 2, 4, 6-10 and 18-26 are rejected under 35 U.S.C. 103 as being unpatentable over Sengupta et al (US 20230063131 A1) hereafter Sengupta, further in view of Shukla et al (US 20210312904 A1) hereafter Shukla, and further in view of Lukyanenko et al (US 20230063713 A1) hereafter Lukyanenko.
With respect to claim 1, Sengupta teaches a method, comprising:
receiving a workflow domain specification associated with at least a portion of eligible steps of a workflow relating to the task, wherein the workflow domain specification comprises, for each respective eligible step of at least the portion of eligible steps, a name of the respective eligible step and a natural language description of the respective eligible step (a virtual agent may select from an action space, which includes a group of primitive actions, an appropriate action. The action space includes 7 categories with a total of 31 primitive actions, wherein the categories are specification. DPL 214 is an example of a dialogue manager. The agent’s primary task is to predict the most appropriate action for a given state in a dialogue system. The method includes receiving first user utterance, identifying a goal based on the input, and selecting an action based on the goal [par. 0018, 0052, 0060]);
determining machine learning model input data based on (i) the content of the completed virtual agent dialog and (ii) the workflow domain specification (a dialogue manager configured to receive a first input based on a first user utterance, and then to identify a first goal based on the first input. An example of user seeking a new mobile device, and the virtual agent (VA) offers a first product based on the user stated specifications [par. 0064-0066]).
However, Sengupta does not disclose processing the machine learning model input data using a trained machine learning model executing on one or more hardware processors to generate a sequence of workflow steps that summarizes the completed virtual agent dialog, wherein the trained machine learning model is conditioned by the workflow domain specification to select one or more workflow steps of the sequence of workflow steps from the eligible steps of the workflow domain specification, and wherein, during training, the trained machine learning model had no exposure to at least one workflow step of the one or more workflow steps; and receiving content of a completed virtual agent dialog that has occurred between at least two communication parties to resolve a task.
In the same field of endeavor, Shukla teaches receiving content of a completed virtual agent dialog that has occurred between at least two communication parties to resolve a task (a dialog correction tool including a logged conversation, a graph-based dialog, an exception notification and an utterance correction window configured to depict a graphical user interface. The conversation between human user and a virtual assistant is logged to the windows that resolves a particular task. The logged conversation sub-window may provide a complete or partial set of utterances associated with the logged conversation. The method may include enabling a user to analyze an exception that occurred during a user-system dialog by reviewing a log dialog and editing the log dialog to create a corrected dialog [par. 0062, 0063, 0072, 0073 and FIG. 5A]).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the invention to have incorporated the concept of improving dialog management for task-oriented dialog systems as suggested by Shukla into the concept of improving a current virtual agent (VA) conversational experiences in real-time with human customers as suggested by Sengupta because both of these systems addressing the process of improving the user experience in interacting with virtual assistance, virtual agents or chatbots. Doing so would be desirable because the system of Sengupta would be more efficient by providing a graphical user interface such as a user-system dialog in real-time that receives user input to communicate with the virtual assistants to solve a particular task (Shukla, [par. 0002-0008]).
However, the combination of Sengupta and Shukla does not particularly disclose processing the machine learning model input data using a trained machine learning model executing on one or more hardware processors to generate a sequence of workflow steps that summarizes the completed virtual agent dialog, wherein the trained machine learning model is conditioned by the workflow domain specification to select one or more workflow steps of the sequence of workflow steps from the eligible steps of the workflow domain specification, and wherein, during training, the trained machine learning model had no exposure to at least one workflow step of the one or more workflow steps.
In the same field of endeavor, Lukyanenko teaches processing the machine learning model input data using a trained machine learning model executing on one or more hardware processors to generate a sequence of workflow steps that summarizes the completed virtual agent dialog (live chat services may include live agents and chatbots for automated help and assistance. A dialogue summary may be provided by determining the most relevant or highest likelihood of interest sentences from dialogue to present to a live agent. This is determined by employing an unsupervised machine learning (ML) model [par. 0009-0014]), wherein the trained machine learning model is conditioned by the workflow domain specification to select one or more workflow steps of the sequence of workflow steps from the eligible steps of the workflow domain specification, and wherein, during training, the trained machine learning model had no exposure to at least one workflow step of the one or more workflow steps (no exposure or zero-shot learning is a paradigm where a model can classify or recognize data it has never seen before during training. The model uses prior knowledge to make predictions. Filtering may include filtering out information noise which may utilize an intent prediction model, such that greetings or approvals of live agents may be recognized by using intent prediction. The dialogue may be split based on messages and the predicted intent of each message. A live chat application may correspond to a computing automation process or device that implements one or more chat workflows and skills for automated assistance. A chat summary may be provided to live agent prior to chat using past dialogue for the user associated with client device [par. 0011-0014, 0035, 0036, 0041-0048]).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the invention to have incorporated the concept of using unsupervised machine learning (ML) for keyword selection and scoring for sentence level dialogue summaries as suggested by Lukyanenko into the combination of Sengupta and Shukla because all of these systems addressing the process of improving the user experience in interacting with virtual assistance, virtual agents or chatbots. Doing so would be desirable because the combination of Sengupta and Shukla would be more efficient by providing a dialogue summary to a live agent prior to chat or during chat with a user when some of the intents are predicted intents based on past dialogue with the users that the ML model may learn to provide a better service when assisting users (Lukyanenko, [par. 0009-0014]).
With respect to claim 2, the combination of Sengupta, Shukla and Lukyanenko teaches wherein the content of the completed virtual agent dialog comprises a plurality of natural language utterances arranged in a sequential time order associated with when the utterances occurred (Shukla, a task-oriented system that help users complete various tasks through natural language conversations, such tasks include customer support or IT helpdesk. Received user input includes user utterances, keyed input, mouse input, etc. [par. 0001, 0004]).
With respect to claim 4, the combination of Sengupta, Shukla and Lukyanenko teaches wherein the at least two communication parties include (i) at least one communication party that is a virtual agent and (ii) zero or more communication parties that are human (Shukla, VA has become increasingly popular in digital world. Various tasks are automated between users and VAs. An application may be a virtual assistant application where a user and a VA exchange requests and responses via a dialog interface [par. 0003, 0020-0023]).
With respect to claim 6, the combination of Sengupta, Shukla and Lukyanenko teaches wherein the task includes a customer support task (Sengupta, VA may be used by corporations to assist customers with tasks such as booking reservations and/or diagnostic issues. VAs or chatbots are designed to assist, support and work with customers to generate alternate solution if the original solution is unavailable [par. 0002, 0024-0026]).
With respect to claim 7, the combination of Sengupta, Shukla and Lukyanenko teaches wherein the workflow domain specification has been selected from a specified list of workflow domain specification options (Sengupta, a specification may be selected from a list of user’s stated specifications. For example, a VA may generate a Samsung Galaxy S5 based on the user’s request of a mobile device and the user’s specifications [par. 0060-0066]).
With respect to claim 8, the combination of Sengupta, Shukla and Lukyanenko teaches wherein the specified list of workflow domain specification options has been determined using a second machine learning model that has been trained to automatically predict a workflow steps domain based on an input dialog (Sengupta, VAs complete their tasks by filling user task constraints and targeting a goal matching the user’s original specifications. The action space includes 7 categories with 31 primitive actions, wherein the categories are specification, request, inform, confirm, result, done and salutation. Depending on the input data, the VAs may generate appropriate actions according to user’s specifications [par. 0027, 0060, 0066]).
With respect to claim 9, the combination of Sengupta, Shukla and Lukyanenko teaches wherein each step of the at least the portion of eligible steps is semantically related to the task (Sengupta, VAs may communicate with human users in a natural language and work with users to perform various tasks, tasks may be referred to goals. A first action is selected based on a goal that the dialogue manager identified based on a first input which is based on a first user utterance. Similarly, a second action is selected based on a second goal. VA may present a second response based on the second action [par. 0002, 0020, 0066]).
With respect to claim 10, the combination of Sengupta, Shukla and Lukyanenko teaches wherein determining the machine learning model input data includes combining (i) the content of the completed virtual agent dialog and (ii) the workflow domain specification according to a specific textual format (Sengupta, a user may be engaged with a VA in a conversation by phone or other voice-based communication devices, and/or through other text-based communication [par. 0039]).
With respect to claim 18, the combination of Sengupta, Shukla and Lukyanenko teaches wherein the sequence of workflow steps comprises a plurality of textual descriptions of actions taken in sequential order to resolve the task (Shukla, the logged conversation sub window may provide a complete conversation between a VA and a user utterances, such that the logged conversation includes a sequential set of user inputs and VA outputs, wherein the outputs provide enough descriptions of actions to be taken by the user [par. 0073, 0077, 0078]).
With respect to claim 19, it is a system that is corresponding to the method of claim 1. Therefore, it is rejected for the same reasons as claimed in claim 1 above.
With respect to claim 20, it is a computer product comprising a non-transitory computer readable medium that is corresponding to the method of claim 1. Therefore, it is rejected for the same reasons as claimed in claim 1 above.
With respect to claim 21, the combination of Sengupta, Shukla and Lukyanenko teaches wherein at least one workflow step of the sequence of workflow steps includes respective parameters that relate to the task (Sengupta, θ represents all the parameters of the function approximator of the RL model. The specification provided defines the eligible steps of the task in a text format, essentially providing the system with a set of parameters for how to complete a task. the system highlights the importance of user task constraints, such as intent, slot, to complete the task. [par. 0037, 0047, 0056]).
With respect to claim 22, the combination of Sengupta, Shukla and Lukyanenko teaches wherein conditioning the trained machine learning model by the workflow domain specification constrains the trained machine learning model to select the one or more workflow steps from the eligible steps of the workflow domain specification (Sengupta, “no exposure” means that a specific piece of information, such as data point or concept or a workflow step, is not included in the dataset used to train the model. A dynamic goal-driven module (GDM) is disclosed to handle conversational paths in a dialogue manager, and goals not included in the initial training data. the virtual agents may be implemented via this GDM system that incorporates policy selection and customer sentiment analysis techniques [par. 0017, 0026, 0031, 0032, 0037, 0045]).
With respect to claim 23, the combination of Sengupta, Shukla and Lukyanenko teaches wherein the sequence of workflow steps comprises an invented step generated by the trained machine learning model, wherein the invented step is not present in the workflow domain specification, and wherein the trained machine learning model is configured to generate a new step name for the invented step (Lukyanenko, the dialogue may be split based on messages and the predicted intent of each message. Chat summarization application includes ML models that may be used for intelligent decision-making and/or predictive outputs. ML models provide such predictive outputs when generating the summaries for the dialogue [par. 0014, 0043, 0044]).
With respect to claim 24, the combination of Sengupta, Shukla and Lukyanenko teaches wherein during training, the trained machine learning model had no exposure to the invented step (Lukyanenko, the dialogue may be split based on messages and the predicted intent of each message. A live chat application may correspond to a computing automation process or device that implements one or more chat workflows and skills for automated assistance. A chat summary may be provided to live agent prior to chat using past dialogue for the user associated with client device [par. 0011-0014, 0035, 0036, 0041-0048]).
With respect to claim 25, the combination of Sengupta, Shukla and Lukyanenko may not teach wherein determining the specified list of specification options comprises: partitioning a dataset of training dialogs into a plurality of batches, wherein each respective batch of the plurality of batches includes semantically similar training dialogs; and processing each respective batch of the plurality of batches using the second machine learning model to predict a corresponding set of steps for each respective batch.
With respect to claim 26, the combination of Sengupta, Shukla and Lukyanenko may not teach wherein determining the specified list of specification options further comprises: determining a combined set of steps by aggregating the corresponding set of steps of each respective batch; and removing duplicates and semantically equivalent steps from the combined set of steps.
Claims 11, 15 and 17 are rejected under 35 U.S.C. 103 as being unpatentable over Sengupta et al (US 20230063131 A1) hereafter Sengupta, further in view of Shukla et al (US 20210312904 A1) hereafter Shukla, as applied in claim 1 above, and further in view of Raffel et al (“Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer”) hereafter Raffel.
Raffel was cited in the IDS filed on 08/01/2022.
With respect to claim 11, the combination of Sengupta and Shukla teaches all limitations as claimed in claim 1 above.
However, the combination of Sengupta and Shukla does not disclose wherein the trained machine learning model is a text-to-text pre- trained language model.
In the same field of endeavor, Raffel teaches wherein the trained machine learning model is a text-to-text pre- trained language model (the procedure for formulating every text processing problem as a text-to-text task. the main utility of transfer learning is the possibility of leveraging pre-trained models in data-scarce settings [page 3, 1. Introduction]).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the invention to have incorporated the concept of transfer learning of a model with pre-trained on a data-rich task before being fine-tuned on a downstream task as suggested by Raffel into the combination of Sengupta and Shukla because all of these systems addressing the process of improving the user experience in interacting with virtual assistance, virtual agents or chatbots by processing text on multi-task learning. Doing so would be desirable because the combination of Sengupta and Shukla would be more efficient by applying transfer learning techniques that for natural language processing (NLP) with introducing a unified framework that converts all text-based language problems into a text-to-text format (Raffel, [page 1, Abstract]).
With respect to claim 15, the combination of Sengupta, Shukla and Raffel teaches wherein the trained machine learning model has been further trained on an additional dataset to perform a summarization task (Raffel, evaluating performance on a wide variety of English-based NLP problems have been conducted, including question answering, document summarization and sentiment classification. An example of CNN/Daily Mail data set was introduced as a question-answering task but was adapted for text summarization [page 2 and 3, 1. Introduction; page 7, 2.3. Downstream Tasks]).
With respect to claim 17, the combination of Sengupta, Shukla and Raffel teaches wherein the trained machine learning model has been further trained to perform a workflow discovery task (Raffel, transfer learning is used to pre-train the model on data-rich tasks, wherein the pre-training causes the model to develop general-purpose abilities and knowledge that can be transferred to downstream tasks (or as a workflow task) [page 1 and 2, 1. Introduction]).
Response to Arguments
The examiner respectfully acknowledges the applicant’s amendments to claims 1, 2, 4, 7, 8, 10 and 19-22.
Applicant’s amendments filed on 12/17/2025 regarding claim rejections under 35 U.S.C. 112(b) to claims 1, 2, 4, 6-11, 15 and 17-26 have been fully considered and are consequently withdrawn.
Applicant’s arguments filed on 12/17/2025 regarding claim rejections under 35 U.S.C. 101 to claims 1, 2, 4, 6-11, 15 and 17-26 have been fully considered but are not persuasive.
Applicant argued that “Applicant submits that the claimed invention integrates any alleged abstract idea into a practical application because it "improves workflow discovery (WD) performance, including in scenarios in which the machine learning model has had no exposure (zero-shot) or little exposure (few-shot) to the types of workflow steps it is expected to extract." (Specification, [0014]).”
Based on what is recited in amended claim 1, the limitations “receiving content of a completed virtual agent dialog that has occurred between at least two communication parties to resolve a task”, “receiving a workflow domain specification associated with at least a portion of eligible steps of a workflow relating to the task …”, “determining machine learning model input data based on (i) the content of the completed virtual agent dialog and (ii) the workflow domain specification” and “generating a representation of the sequence of workflow steps” merely recite mental processes including observation, evaluation, judgement, etc. A human could review a dialog, compare it to a workflow definition, determine applicable workflow steps and/or generate a summary. These limitations can be characterized as evaluating information, classifying information, and/or creating a summary, outputting the result of the analysis.
The limitations “processing the machine learning model input data using a trained machine learning model executing on one or more hardware processors … wherein the trained machine learning model is conditioned by the workflow domain specification … the trained machine learning model had no exposure to at least one workflow step of the one or more workflow steps” are considered as additional elements, as the claim does not explain any ML technology improvement, any computer functionality improvement, or any particular machine transformation.
Applicant pointed to a disclosure in the Specification describing improvements in workflow discovery ad zero-shot prediction capabilities. However, improvements described in the Specification are not sufficient by themselves to establish integration into a practical application. The claim must recite the technological features that produce the alleged improvement. In here, amended claim 1 does not recite a particular model architecture, conditioning technique, embedding technique, training procedure or other technological mechanism that improves ML technology. Instead, the claim broadly recites using a trained ML model to analyze dialog information and generate workflow step summary. Accordingly, the claim does not integrate the judicial exception into a practical application.
Therefore, amended claim 1 and its corresponding claims 19 and 20 are not patent-eligible for at least the reasons listed above. Dependent claims 2, 4, 6-11, 15, 17, 18 and 21-26, those are depended on claim 1, are not patent-eligible for the same reasons.
Applicant’s arguments filed on 12/17/2025 regarding claim rejections under 35 U.S.C. 103 to claims 1, 2, 4, 6-11, 15 and 17-26 have been fully considered and moot in view of the new ground of rejection (see rejection above).
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Koneru et al (US 12340795 B2) disclosed the development and design of bot interfaces and, more particularly, to one or more components, systems and methods of an intelligent development and design platform configured to assist users in the design, development and deployment of bot applications.
Gangireddy et al (US 20230289167 A1) disclosed a method includes: receiving, by a computing device, a plurality of conversation transcripts generated from a plurality of versions of a chatbot; determining, by the computing device, a plurality of changes of a plurality of attributes of intents between the plurality of the versions of the chatbot; identifying, by the computing device, at least one intent to update from the plurality of changes of the plurality of attributes of the intents to improve performance of the chatbot; and generating, by the computing device, another version of the chatbot that includes the at least one intent updated from the plurality of changes of the plurality of attributes of the intents to improve the performance of the chatbot.
Menon et al (US 20230188480 A1) disclosed a method and system for generating and correcting chatbot responses based on reinforcement learning (RL) are disclosed. In some embodiments, the method includes receiving user data associated with a user in a chatbot conversation. The method includes providing a first recommendation to the user. The method includes detecting user feedback to the first recommendation in the chatbot conversation. The method then includes determining whether to assign a positive reward or a negative reward to the user feedback based on sentiment analysis performed on the user feedback.
Makki Niri et al (US 20220414344 A1) disclosed methods and systems for training an intent classifier. For example, a question-intent tuple dataset comprising data samples is received. Each data sample has a question, an intent, and a task. A pre-trained language model is also received and fine-tuned by adjusting values of learnable parameters. Parameter adjustment is performed by generating a plurality of neural network models. Each neural network model is trained to predict at least one intent of the respective question having a same task value of the tasks of the question-intent tuple dataset. Each task represents a source of the question and the respective intent.
Asthana et al (US 20210097110 A1) disclosed a chatbot application is in a mode of live chat conversation with a user, a trained intent classifier determines an intent that underlies a first live question received by the trained intent classifier from the user. A trained next predictor receives the intent from the intent classifier. The trained next intent predictor generates a set of predicted next intents responsive to receiving the intent. A trained re-ranker selects at least one of the predicted next intents responsive to receiving the set of predicted next intents. A question selection engine sends at least one suggested question to the user responsive to receiving the at least one predicted next intent.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to Quoc Phung whose telephone number is (703) 756 1330. The examiner can normally be reached on Monday through Friday from 9am to 5pm PT.
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, Jennifer Welch can be reached on 571-272-7212. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of an application may be obtained from the Patent Application Information Retrieval (PAIR) system. Status information for published applications may be obtained from either Private PAIR or Public PAIR. Status information for unpublished applications is available through Private PAIR only. For more information about the PAIR system, see http://pair-direct.uspto.gov. Should you have questions on access to the Private PAIR system, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative or access to the automated information system, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/Q.L.P./Examiner, Art Unit 2143
/JENNIFER N WELCH/Supervisory Patent Examiner, Art Unit 2143
Prosecution Insights