DETAILED ACTION
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 .
Acknowledgments
The submission filed on 03/10/26 is acknowledged.
Status of Claims
Claims 1-11, 13 and 15-22 are pending.
In the Amendment filed on 03/10/26, claims 1, 4-11, 13 and 15-20 were amended, claims 12 and 14 were cancelled, and claims 21 and 22 were added.
Claims 1-11, 13 and 15-22 are rejected.
Response to Arguments
Regarding the rejections under 35 U.S.C. 112
The previous rejection is moot in view of the cancellation of claim 12. However, the instant amendments give rise to new rejections, as set forth and explained in the body of the Office Action hereinbelow.
Regarding the rejection under 35 U.S.C. 101
Applicant's arguments have been fully considered but are not persuasive.
The Office responds to Applicant's arguments below.
Applicant argues that "the claims are not directed to any abstract idea," specifically, that
There is no explanation why the claims are considered to fall into any of these concepts. For example, the claims do not describe any fundamental economic principle …. (Response, p. 9)
The Office respectfully disagrees.
The explanation in question has been provided. First, as per the rejection, the claims "recite a method, system, or machine-readable storage device for training a customer service agent" (see Claim Rejections - 35 USC § 101, in previous Office Action and hereinbelow). Second, the previous/instant rejection explained/explains, e.g., that the bolded content of claim 1 is an abstract idea, e.g., the limitations of establishing a communication session with .. a listing network …; generating, … a conversation segment based on historical user interaction representing a customer support issue associated with the listing network …; transmitting, … the conversation segment … via the communication session; receiving one or more responses to the conversation segment … in the communication session; … a simulated conversation comprising at least the conversation segment generated … and the one or more responses received …; based on determining that a difficulty level associated with the simulated conversation transgresses a difficulty threshold, selecting a … judge … from a plurality; generating a score for the simulated conversation using the … judge … that was selected from the plurality …; and causing parameters … to be updated based on the score for the simulated conversation transgressing a threshold, describe/encompass the abstract idea of training a customer service agent, which is a fundamental economic practice or principle and/or a commercial or legal interaction. For example, training a customer service agent is comparable to "structuring a sales force or marketing company, which pertains to marketing or sales activities or behaviors," which is a commercial or legal interaction. MPEP 2106.04(a)(2)II.B.
Applicant further argues that "the claims are directed to a technical solution to a technical problem," citing portions of specification paragraphs 0015 and 0016 as support for this assertion. Response, pp. 9-10.
Paragraphs 0015 and 0016 of the specification are set forth below.
[0015] Certain systems allow users to chat with virtual agents or bots to find resolutions to the issues. However, these agents are not sophisticated enough to be able to provide resolutions to most issues as they generally are programmed to provide static responses to queries that include certain keywords. As such, even accessing such virtual agents can waste time and frustrate users and the users may still have to contact a live human agent to resolve issues. Such repetitive and manual processes are incredibly time-consuming and can be very frustrating to end users. This can result in missed opportunities and wasted computational resources. In addition, training these virtual agents to respond in a meaningful way to customer support tickets (e.g., customer issues) relies on a vast amount of training conversations and labeling. Generating such simulated conversations is also incredibly time-consuming and still results in training conversations that lack any diversity. This causes the virtual agents to be trained to provide unrealistic responses and also reduces the array of issues that the virtual agents are programmed to handle.
[0016] To address these technical problems, the disclosed techniques provide a network site that allows a first machine learning model to interact in a communication session with an agent (e.g., a virtual agent or human agent) on the listing network site to generate a simulated conversation. The first machine learning model can simulate a real-world customer based on prior customer interactions in a vast array of settings and personalities. Multiple of these simulated conversations between the first machine learning model and the agent are stored and then analyzed to train or update parameters of the virtual agent. This improves the quality of responses and vast array of solutions that the virtual agent is capable of generating. In some cases, the simulated conversations are scored (manually by a human) and/or automatically by a second machine learning model (e.g., a virtual judge). These scores are used to guide the updating of the parameters to improve the overall functioning of the virtual agent and the device.
As seen from paragraphs 0015 and 0016 above, the solution provided by Applicant is improving the quality of responses by training a virtual agent, by having simulated interactions with a customer -- simulated conversations -- with a variety of settings an personalities, and analyzing the conversations to train or update parameters of the agent. As such, the solution provided by Applicant is a putative improvement to the abstract idea, not an improvement in computer functioning or other technology. The benefits flowing from the solution, including eliminating or reducing "missed opportunities and wasted computational resources," are merely the natural result of the putative improvement to the abstract idea (improved quality of responses); they do not stem from any improvement in computer functioning/other technology. The non-abstract idea elements of the claims are simply off-the-shelf elements used in their ordinary capacities/functionalities, and they are recited at a high level of generality and not described.
Subject Matter Distinguishable From Prior Art
The cited prior art of record, either alone or in combination, fails to expressly teach or suggest the features found in independent claims 1, 18 and 20.
Klein (US-20250182128-A1) teaches a system that generates relevant and vetted training data using intelligent simulated users and evaluation of conversation data. A simulated user and an automated agent engage in a conversation to generate conversation and/or interaction data. The simulated user is guided by scenarios which are generated based on one or more controls to be followed by the automated agent. Using a simulated user driven by control-derived scenarios ensures the ensuing conversation data is relevant to the desired scope of operation for the automated agent. The conversation data is evaluated based on the controls to confirm the automated agent actions and responses followed the controls properly. Evaluating the conversation data based on the controls ensures that conversation data associated with properly followed controls is used as subsequent training data.
Wooters (US-20250029110-A1) teaches a method of evaluating performance of a chatbot including identifying a test scenario including a request and an expected outcome, initiating an automated conversation between a first machine learning model and the chatbot based on the test scenario, storing a recording of the automated conversation, providing the recording of the automated conversation and the test scenario to a second machine learning model, and receiving an evaluation of the automated conversation from the second machine learning model based on the recording of the automated conversation and the expected outcome.
Sundaram (US-2025/0217224-A1) teaches training a customer support agent for responding to user queries re installation, diagnostics and debugging of complex systems, including receiving a natural language (NL) query, providing a prompt based on the NL query to a language model (LM), and receiving from the LM a response to the NL query; training the LM using logs of actual historical problems and synthetic logs generated specifically for training purposes; supervised training comprising evaluating LM-generated training responses on a scale measuring the difference between the LM's training responses and ground truth responses, where the LM attempts to emulate the ground truth responses; unsupervised training using historical user queries and respective expert responses; and obtaining an evaluation score characterizing effectiveness of the training NL response, and modifying the parameters of the LM based on the evaluation score. Sundaram also teaches the use of CNNs in relevant context.
Caron (US-20210127004-A1) teaches generating a simulated caller dialog including a caller intended issue for a scenario for testing a customer service representative (CSR), generating an intent determination recognition score and a CSR score and various subscores, and changing the behavior of the CSR based on the scores.
Beaver (US-20240355318-A1) teaches training an intelligent virtual assistant based on transcripts of historical interactions between customers and customer service agents, including reinforcement learning based on received customer input and similarity between a response provided by a customer service agent and a predicted response generated by the second language model.
Akkiraju (US-20190109802-A1) teaches training a chatbot for a customer chat simulation based on a customer service conversation data, a task scenario, and a customer persona, determining an assessment of the performance of the customer service agent based on the interaction between the customer service agent and the chatbot, and generating feedback for the customer service agent based on the assessment of the performance of the customer service agent.
Pauls (US-20250131202-A1) teaches a system for providing and managing an automated agent that may interact with a customer and utilize rules and instructions to determine a response to the customer and actions to perform, including a machine learning model, which may be implemented as a large language model (LLM).
Earle (US-20240111960-A1) teaches scoring a simulated conversation based on criteria and, based on the score, adjusting parameters, such as weights of a customer module or agent module.
Jang (KR-20250098341-A) (filed 12/22/23) teaches that if a difficulty level of a conversation exceeds a threshold, a computing device processing natural language may utilize an additional artificial intelligence model, which may include, e.g., an additional LLM (Large Language Model).
Copeland (US-20230267370-A1) (0068) teaches that an intelligent assistant may use a rules engine to generate a follow up action or communication where a complexity level of the interactive situation does not exceed a threshold, and may use an ML where a complexity level of the interactive situation exceeds a threshold.
Deegan (US-20230353675-A1) (0018) teaches that in a customer support scenario, a customer's message may be first routed to the chatbot, and the chatbot may have a confidence level associated with an automatically generated answer with respect to the chatbot's ability to accurately answer the customer's question; if the confidence level of the answer is below a predetermined threshold, one or both of the user's communication and the chatbot's proposed answer may be sent by the chat router to a live agent.
Emery (US-20180181558-A1) teaches a method for evaluating user-machine conversation, wherein if a bot's reply to a user's query is not valid or is of low confidence (e.g., if the user changes the topic to a topic the bot is not competent to answer about), the system switches to a different bot to provide a reply.
Tuma (US-20210064860-A1) teaches a method of extracting text for a document, including selecting a particular machine learning model in response to determining that a level of complexity (in identifying information to be extracted from a document) exceeds a threshold.
Leong (US-20220366896-A1) teaches a method of training a trainee skills by a bot, including selecting a different response or a different algorithm based on a level of difficulty at which the trainee is being trained.
However, in particular, the cited prior art of record, either alone or in combination, fails to expressly teach or suggest all of the features in independent claims 1, 18 or 20 and more specifically the limitations of: based on determining that a difficulty level associated with the simulated conversation transgresses a difficulty threshold, selecting a virtual judge machine learning model from a plurality of models comprising a first model and a second model; generating a score for the simulated conversation using the virtual judge machine learning model that was selected from the plurality of models; and
causing parameters of the virtual bot to be updated based on the score for the simulated conversation transgressing a threshold, in combination with the other claim limitations.
Claim Objections
Claims 1, 10, 18 and 20 are objected to because of the following informalities:
A) Claims 1, 18 and 20 recite:
A … for generating simulated conversations between a machine learning model and a virtual bot to improve functioning of the virtual bot … comprising:
[…]
establishing a communication session with the virtual bot of a listing network platform;
It is not clear whether or not "the virtual bot of a listing network platform" (hereafter, "the latter recitation") refers back to the previous recitations of "a virtual bot" and "the virtual bot" (in other words, it is not clear whether or not "the virtual bot of a listing network platform" is the same virtual bot as the previous recitations of "a virtual bot" and "the virtual bot"), because the latter recitation recites that the virtual bot is "of a listing network" while the previous recitations did not recite that the virtual bot is "of a listing network." The recitations in question are understood to encompass a typographical/ clerical error. Applicant is understood to have intended that the latter recitation of the virtual bot does indeed refer back to the previous recitations of "a virtual bot" and "the virtual bot" (in other words, Applicant is understood to have intended that "the virtual bot of a listing network platform" is the same virtual bot as the previous recitations of "a virtual bot" and "the virtual bot.") For purposes of examination, the language in question is interpreted according to this understanding.
B) Claim 10 recites:
wherein generating a score is based on one or more criteria.
Base claim 1, from which claim 10 depends, recites:
generating a score for the simulated conversation using the virtual judge machine learning model that was selected from the plurality of models; and
It is not clear whether or not the recitation of "generating a score" in claim 10 refers back to the recitation of "generating a score" in claim 1. The recitation of claim 10
is understood to be a typographical/clerical error. Applicant is understood to have intended that the recitation of "generating a score" in claim 10 does indeed refer back to the recitation of "generating a score" in claim 1. If this understanding is correct, then Applicant should use the definite article ("the") to indicate that the recitation of "generating a score" in claim 10 refers back to the recitation of "generating a score" in claim 1. For purposes of examination, the recitation of "generating a score" in claim 10 is interpreted as referring back to the recitation of "generating a score" in claim 1.
Appropriate correction is required.
Claim Rejections - 35 U.S.C. § 11235 U.S.C. § 112(a)
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 1-11, 13 and 15-22 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 pre-AIA the inventor(s), at the time the application was filed, had possession of the claimed invention.
Lack of Written Description/Not in Specification
Claims 1, 18 and 20 recite:
based on determining that a difficulty level associated with the simulated conversation transgresses a difficulty threshold, selecting a virtual judge machine learning model from a plurality of models comprising a first model and a second model;
Support in the disclosure is not found for the above-indicated recitation.
Before turning to the disclosure, we present the following background to the issue at hand.
Prior to the instant claim amendments, independent claims 1, 18 and 20 recited:
determining whether the difficulty level associated with the simulated conversation transgresses a difficulty threshold;
based on a result of determining whether the difficulty level associated with the simulated conversation transgresses the difficulty threshold, selecting a virtual judge scoring model from a plurality of models comprising a first model and a second model;
and dependent claims 2 and 3 recited:1
wherein the first model comprises a large language model (LLM) and is selected in response to determining that the difficulty level fails to transgress the difficulty threshold. (claim 2)
wherein the second model comprises a convolutional neural network (CNN) and is selected in response to determining that the difficulty level transgresses the difficulty threshold. (claim 3)
Thus, to summarize, prior to the instant claim amendments, claim 1 recited determining whether the difficulty level transgressed the threshold, and then based on that determination one of two models is selected: specifically, per claim 2, based on a determination that the difficulty level transgressed the threshold, an LLM is selected, and per claim 3, based on a determination that the difficulty level did not transgress the threshold, a CNN is selected. Put another way, claim 1 recited a fork, and claims 2 and 3 recited exclusive outcomes of the respective two prongs of the fork: once a given prong of the fork is selected (i.e., once it is determined whether the threshold was transgressed or not), then the outcome (i.e., which model is selected) follows directly from the selection of the given prong.
As best understood, Applicant's specification is consistent with this subject matter of the previous versions of claims 1-3, i.e., as best understood, Applicant's specification is consistent with the content set forth in the preceding paragraph.
The instant claim amendments are at variance with this subject matter of the previous versions of claims 1-3. The instant amendments deny that, once a given prong of the fork is selected (i.e., once it is determined whether the threshold was transgressed or not), then the outcome (i.e., which model is selected) follows directly from the selection of the given prong. Specifically, claim 1 as currently amended recites that, based on a determination that the difficulty level transgressed the threshold, (once a given prong of the fork is selected), either of the two models is selected (either outcome can follow).
As best understood, Applicant's specification is not consistent with this subject matter of the current version of claim 1.
We turn now to Applicant's disclosure as filed to address whether the recitation in question is supported.
As for the subject matter of the recitation in question, the "virtual judge" is discussed in the disclosure primarily at paragraphs 0044-0049 of the specification, with additional brief mentions at paragraphs 0016 and 0067 of the specification and in original claim 12. The "difficulty level" is discussed solely at 0049 of the specification. 0049 reads as follows:
In some cases, a difficulty level can be assigned or determined for a particular simulated conversation. The virtual judge 620 determines if the difficulty level transgresses a difficulty threshold. [1] In response to determining that the difficulty level transgresses the difficulty threshold, the virtual judge 620 uses the CNN to generate the score for the simulated conversation. [2] In some cases, in response to determining that the difficulty level fails to transgress the difficulty threshold, the virtual judge 620 uses an LLM to generate the score for the simulated conversation. In some cases, the virtual judge 620 uses the CNN to generate a first score associated with a first criterion and uses the LLM to generate a second score associated with a second criterion for an individual simulated conversation. In this way, the score 630 generated for the simulated conversation can be made up of scores generated by different system components (e.g., a CNN, an LLM, and/or a human). (Emphasis added; bracketed numerals added to facilitate discussion below)
As per 0049:
[1] In response to determining that the difficulty level transgresses the difficulty threshold, the virtual judge 620 uses the CNN to generate the score for the simulated conversation.
That is, 0049 states that if the difficulty level transgresses the difficulty threshold, then the CNN is used (selected), period. 0049 does not state that if the difficulty level transgresses the difficulty threshold, then a selection is made of which of two models (CNN and LLM) to use. The single sentence [1] of 0049 quoted above does not appear to give leeway to use another model (i.e., a model other than the CNN) in the scenario in which the difficulty level transgresses the threshold.
Note further, as per 0049:
[2] In some cases, in response to determining that the difficulty level fails to transgress the difficulty threshold, the virtual judge 620 uses an LLM to generate the score for the simulated conversation.
That is, 0049 states that, in some cases, if the difficulty level does not transgress the difficulty threshold, then the LLM is used (selected). Unlike the case where the difficulty level transgresses the difficulty threshold, here the single sentence [2] of 0049 quoted above does give leeway to use another model (i.e., a model other than the LLM) in the scenario in which the difficulty level does not transgress the threshold.
Given that the drafter of the application explicitly provided leeway -- by adding the words "in some cases" -- to use either model in the scenario in which the difficulty level does not transgress the threshold, the reader understands that the drafter was clearly aware of the possibility of providing this leeway or not, when the drafter set forth the scenario in which the difficulty level does transgress the threshold. That is, having used the phrase "in some cases" in sentence [2], the drafter clearly had in mind the possibility of using this same phrase in adjacent sentence [1]. And yet the drafter explicitly failed to use this phrase in adjacent sentence [1] -- explicitly failed to provide leeway for the scenario in which the difficulty level transgresses the threshold. This suggests that the drafter knowingly and intentionally omitted the phrase "in some cases" in sentence [1], i.e., that the drafter's failure to provide the leeway for this scenario was knowing and intentional.
Thus, the specification appears to teach that if the difficulty level transgresses the threshold, no choice of selecting a model is provided but rather the CNN is necessarily used, whereas if the difficulty level does not transgress the threshold, a choice of selecting a model is provided, i.e., the LLM may be used or not, and presumably, in light of the context, either the LMM or the CNN may be used.
Further, if 0049 be deemed ambiguous/inconclusive as to supporting the foregoing reasoning and conclusion, the principle of contra proferentem2 would appear to bolster the foregoing reasoning and conclusion. That is, the drafter had total control over how to write the application. Accordingly, the burden of ambiguity should be held against the drafter, not against the public, which stands to lose from Applicant's potential future patent monopoly rights to claims unsupported by the disclosure.
---
The end part of 0049, that is, the part following sentence [2], bears commenting on.
The end part of 0049 reads as follows:
[3] In some cases, the virtual judge 620 uses the CNN to generate a first score associated with a first criterion and uses the LLM to generate a second score associated with a second criterion for an individual simulated conversation. [4] In this way, the score 630 generated for the simulated conversation can be made up of scores generated by different system components (e.g., a CNN, an LLM, and/or a human). (Bracketed numerals added to facilitate discussion below)
As best understood, this part of 0049 should have been presented as a separate paragraph from the preceding part of 0049, because, as best understood, this part of 0049 is no longer directed to the topic of a difficulty level, but rather is directed to the topic of using a criterion in generating the score. As such, this part of 0049 relates to and may be understood as a continuation of the discussion of this topic in 0044. As best understood, the reason why this part of 0049 is presented in 0049 is because it also relates to using both a CNN and an LLM to generate two scores, respectively.
Let us turn now to the question of how the end part of 0049 bears on the question of support for the limitation of claim 1 in question.
The sentence [3] might be taken to suggest that, in the scenario in which the difficulty level transgresses the threshold, both the CNN and the LLM could be used, and therefore, either one could be used and either one could be selected. However, this suggestion appears to contradict sentence [1], which does not appear to provide such leeway, as explained above.
Note then that the sentence [3] starts with the phrase "in some cases." The plain meaning of "in some cases" is that it encompasses the meaning of "not in all cases." It could be that the phrase "in some cases" in sentence [3] is intended to mean that sentence [3] applies to the scenario in which the difficulty level does not transgress the threshold (i.e., sentence [2]), but does not apply to the scenario in which the difficulty level transgresses the threshold (i.e., sentence [1]). If this were the case, then it would reconcile the indicated contradiction (i.e., between sentences [3] and [1]), and sentence [1] would retain its full force, i.e., to the effect that in the scenario in which the difficulty level transgresses the threshold, only the CNN is used.
In any event, the end part of 0049 does not explicitly say that it applies to (-- does not explicitly say that both models or either model could be used in --) the embodiment involving a difficulty level. In addition, the end part of 0049 is directed to an embodiment characterized by the use of the two models together, not to an embodiment characterized by selecting a single one of the models from among the two models.
In the end, the drafter's intention in the end part of 0049 is not entirely clear. But on balance, this apparent unclarity/ambiguity does not appear to defeat the reasoning and conclusion set forth above pertaining to the preceding part of 0049.
Accordingly, support in the disclosure is not found for the above-indicated recitation of claims 1, 18 and 20.
Claims 2-11, 13, 15-17, 19 and 21-22 are rejected by virtue of their dependency from a rejected claim.
35 U.S.C. § 112(b)
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.
Claims 2 and 3 are 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 pre-AIA the applicant regards as the invention.
Unclear Antecedent Basis/Unclear Scope
Claim 2 recites:
wherein the first model comprises a large language model (LLM) and is selected in response to determining that the difficulty level fails to transgress the difficulty threshold.
Base claim 1 recites:
based on determining that a difficulty level associated with the simulated conversation transgresses a difficulty threshold, selecting a virtual judge machine learning model from a plurality of models comprising a first model and a second model;
As per the underlined language above, claim 2 appears to contradict claim 1. Claim 1 recites "determining that a difficulty level associated with the simulated conversation transgresses a difficulty threshold." In contrast, claim 2 recites "determining that the difficulty level fails to transgress the difficulty threshold." Note that "the difficulty level" in claim 2 refers to "the difficulty level" in claim 1, which is associated with a particular conversation; therefore, "the difficulty level" in claim 2 pertains to the same particular conversation as in claim 1. Logically, it does not seem possible for a difficulty level of a given conversation to both (be determined to) transgress a difficulty threshold (as per claim 1) and (be determined to) fail to transgress the same difficulty threshold (as per claim 2). Logically, this scenario amounts to saying that something is both X and not X. While one might imagine scenarios in which something could both be X and not X (e.g., at different times, from different perspectives, or based on different measurement scales), Applicant's disclosure does not offer any account whatsoever of such possibilities. (In any event, the idea of different times does not seem to be applicable here, as the difficulty level of a given conversation would not appear to change over the time scales that would appear to pertain to the disclosure.) Therefore, absent such extenuating circumstances in the disclosure, claim 2 appears to contradict claim 1 by virtue of logic. As such, the meaning, and hence the scope, of claim 2 is not clear. MPEP 2173.03 ("In addition, inconsistencies in the meaning of terms or phrases between claims may render the scope of the claims to be uncertain. Tvngo Ltd. (BVI) v. LG Elecs. Inc., 861 Fed. Appx. 453, 459-60, 2021 USPQ2d 697 (Fed. Cir. 2021) ("The issue is not breadth of the dependent claims but their use of the disputed phrase in a way that contradicts the independent claims. The dependent claims state that 'said overlay activation criterion includes . . . a user command information,' which conflicts with the independent claim's use of this same phrase."). "When faced with this unknown and undefined phrase, a skilled artisan would look for clarification not only in the specification but also in '[o]ther claims of the patent in question,' which 'can also be valuable sources of enlightenment as to the meaning of a claim term.'" Id. at 460 (quoting Philips v. AWH Corp., 415 F.3d 1303, 1314, 75 USPQ2d 1321, 1327 (Fed. Cir. 2005)).")
Claim 3 is rejected by virtue of its dependency from a rejected claim.
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, 13 and 15-22 are rejected under 35 U.S.C. 101 because the claimed invention is directed to a judicial exception (i.e., a law of nature, a natural phenomenon, or an abstract idea) without significantly more.
Claims 1-11, 13 and 15-22 are directed to a method, system, or machine-readable storage device, which are/is one of the statutory categories of invention. (Step 1: YES)
Claims 1, 18 and 20 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more. The claims recite a method, system, or machine-readable storage device for training a customer service agent.
For claims 1, 18 and 20 (claim 1 being deemed representative), the limitations (indicated below in bold) of:
establishing a communication session with the virtual bot of a listing network platform;
generating, by the machine learning model, a conversation segment based on historical user interaction representing a customer support issue associated with the listing network platform;
transmitting, by the machine learning model, the conversation segment to the virtual bot via the communication session;
receiving one or more responses to the conversation segment from the virtual bot in the communication session;
storing a simulated conversation comprising at least the conversation segment generated by the machine learning model and the one or more responses received from the virtual bot ;
based on determining that a difficulty level associated with the simulated conversation transgresses a difficulty threshold, selecting a virtual judge machine learning model from a plurality of models comprising a first model and a second model;
generating a score for the simulated conversation using the virtual judge machine learning model that was selected from the plurality of models; and
causing parameters of the virtual bot to be updated based on the score for the simulated conversation transgressing a threshold.
as drafted, constitute a process that, under the broadest reasonable interpretation, covers "certain methods of organizing human activity," specifically, "fundamental economic practices or principles" and/or "commercial or legal interactions," but for recitation of generic computer components. The Examiner notes that "fundamental economic practices" or "fundamental economic principles" describe concepts relating to the economy and commerce, including hedging, insurance, and mitigating risks, and "commercial interactions" or "legal interactions" include agreements in the form of contracts, legal obligations, advertising, marketing or sales activities or behaviors, and business relations. MPEP 2106.04(a)(2)II.A.,B. If a claim limitation, under its broadest reasonable interpretation, covers "fundamental economic practices or principles" and/or "commercial or legal interactions," but for recitation of generic computer components, then it falls within the "certain methods of organizing human activity" grouping of abstract ideas. Accordingly, claims 1, 18 and 20 recite an abstract idea. (Step 2A - Prong 1: YES. The claims recite an abstract idea.)
This judicial exception is not integrated into a practical application. Claims 1, 18 and 20 recite the additional elements of simulated, a machine learning model, a virtual bot, a (listing network) platform, storing, virtual (judge machine learning model), of models comprising a first model and a second model (the foregoing recited by claims 1, 18 and 20), one or more processors of a machine, a memory storing instructions that, when executed by the one or more processors, cause the machine to perform operations comprising: (the foregoing additionally recited by claim 18), and a machine-readable storage device embodying instructions … that, when executed by a machine, cause the machine to perform operations comprising: (the foregoing additionally recited by claim 20), that implement the abstract idea. These additional elements are not described by the applicant and they are recited at a high level of generality (i.e., one or more generic computer elements performing generic computer functions), such that they amount to no more than mere instructions to apply the exception using generic computer elements. Accordingly, even in combination 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. (Step 2A - prong 2: NO. The additional elements do not integrate the abstract idea into a practical application.)
The claims do not include additional elements that are sufficient to amount to significantly more than the judicial exception itself. As discussed above with respect to integration of the abstract idea into a practical application, the additional elements of simulated, a machine learning model, a virtual bot, a (listing network) platform, storing, virtual (judge machine learning model), of models comprising a first model and a second model (the foregoing recited by claims 1, 18 and 20), one or more processors of a machine, a memory storing instructions that, when executed by the one or more processors, cause the machine to perform operations comprising: (the foregoing additionally recited by claim 18), and a machine-readable storage device embodying instructions … that, when executed by a machine, cause the machine to perform operations comprising: (the foregoing additionally recited by claim 20), to perform the noted steps amount to no more than mere instructions to apply the exception using generic computer elements. Mere instructions to apply an exception using generic computer elements cannot provide an inventive concept ("significantly more"). Accordingly, even in combination, these additional elements do not provide significantly more. As such, claims 1, 18 and 20 are not patent eligible. (Step 2B: NO. The claims do not provide significantly more.)
Dependent claims 2-11, 13, 15-17, 19, 21 and 22 are similarly rejected because they further define/narrow the abstract idea of independent claims 1, 18 and 20 as discussed above, and/or do not integrate the abstract idea into a practical application or provide an inventive concept such as would render the claims eligible, whether each is considered individually or as an ordered combination.
As for further defining/narrowing the abstract idea:
Dependent claims 2-11, 13, 15-17, 19, 21 and 22 merely describe being selected in response to determining whether the difficulty level transgresses the difficulty threshold (claims 2 and 3); generating scores (claim 4); simulating a customer (claim 5); wherein the conversation segment corresponds to an individual customer personality of a plurality of customer personalities that are being represented (claim 6); accessing a plurality of historical customer support tickets comprising a plurality of historical conversations (of) a plurality of users; and based on the plurality of historical customer support tickets to generate the conversation segment representing the customer support issue associated with the listing network (claim 7); generating a prompt comprising an instruction to leverage the plurality of historical customer support tickets to generate the conversation segment representing the customer support issue associated with the listing network, the customer support issue representing at least one customer support issue specified in at least one of the plurality of historical conversations (claim 8); wherein the prompt comprises a type of personality of a plurality of personalities to use in order to control a tone associated with the conversation segment (claim 9); wherein generating a score is based on one or more criteria (claim 10); wherein the parameters are updated based on the one or more criteria (claim 11); judge; accessing the one or more criteria, the one or more criteria being associated with instructions for assigning a score to the one or more criteria, accessing one or more training scores generated, using the one or more criteria, for one or more training conversations (with) one or more users, and generating a prompt with an instruction to generate the score based on the one or more criteria, the one or more training conversations, and the one or more training scores (claims 13 and 16); judge; accessing training data comprising a plurality of training conversations (of) one or more users and corresponding ground truth training scores, processing the training data to predict a training score for an individual training conversation of the plurality of training conversations, computing a deviation between the training score and ground truth training score corresponding to the individual training conversation, and updating one or more parameters based on the computed deviation (claims 15 and 17); and judge (claim 19).
As for additional elements:
Claim 2 recites “wherein the first model comprises a large language model (LLM).” This recitation is at a high level of generality such that it amounts to no more than mere instructions to apply the exception using a generic computer element. Even in combination these additional elements do not integrate the abstract idea into a practical application and do not amount to significantly more than the abstract idea itself.
Claim 3 recites “wherein the second model comprises a convolutional neural network (CNN)." This recitation is at a high level of generality such that it amounts to no more than mere instructions to apply the exception using a generic computer element. Even in combination these additional elements do not integrate the abstract idea into a practical application and do not amount to significantly more than the abstract idea itself.
Claim 4 recites "using/by the first model" and "using/by the second model." This recitation is at a high level of generality such that it amounts to no more than mere instructions to apply the exception using a generic computer element. Even in combination these additional elements do not integrate the abstract idea into a practical application and do not amount to significantly more than the abstract idea itself.
Claim 5 recites “wherein the first machine learning model is trained" and "a virtual" (customer). This recitation is at a high level of generality such that it amounts to no more than mere instructions to apply the exception using a generic computer element. Even in combination these additional elements do not integrate the abstract idea into a practical application and do not amount to significantly more than the abstract idea itself.
Claim 6 recites “the machine learning model is trained." This recitation is at a high level of generality such that it amounts to no more than mere instructions to apply the exception using a generic computer element. Even in combination these additional elements do not integrate the abstract idea into a practical application and do not amount to significantly more than the abstract idea itself.
Claim 7 recites “wherein the first machine learning model comprises a large language model (LLM),” "the virtual bot," "training the LLM," and the "platform." This recitation is at a high level of generality such that it amounts to no more than mere instructions to apply the exception using a generic computer element. Even in combination these additional elements do not integrate the abstract idea into a practical application and do not amount to significantly more than the abstract idea itself.
Claim 8 recites “the LLM” and the "platform." This recitation is at a high level of generality such that it amounts to no more than mere instructions to apply the exception using a generic computer element. Even in combination these additional elements do not integrate the abstract idea into a practical application and do not amount to significantly more than the abstract idea itself.
Claim 11 recites "the virtual bot." This recitation is at a high level of generality such that it amounts to no more than mere instructions to apply the exception using a generic computer element. Even in combination these additional elements do not integrate the abstract idea into a practical application and do not amount to significantly more than the abstract idea itself.
Claim 13 recites “wherein the virtual … machine learning model comprises a large language model (LLM),” "the virtual bot," and "the LLM." This recitation is at a high level of generality such that it amounts to no more than mere instructions to apply the exception using a generic computer element. Even in combination these additional elements do not integrate the abstract idea into a practical application and do not amount to significantly more than the abstract idea itself.
Claim 15 recites “wherein the virtual … machine learning model comprises a convolutional neural network (CNN),” “training the CNN by performing training operations," "the virtual bot," and "the CNN." This recitation is at a high level of generality such that it amounts to no more than mere instructions to apply the exception using a generic computer element. Even in combination these additional elements do not integrate the abstract idea into a practical application and do not amount to significantly more than the abstract idea itself.
Claim 16 recites "the virtual bot" and “a large language model (LLM)." This recitation is at a high level of generality such that it amounts to no more than mere instructions to apply the exception using a generic computer element. Even in combination these additional elements do not integrate the abstract idea into a practical application and do not amount to significantly more than the abstract idea itself.
Claim 17 recites "the virtual bot" and "a/the neural network." This recitation is at a high level of generality such that it amounts to no more than mere instructions to apply the exception using a generic computer element. Even in combination these additional elements do not integrate the abstract idea into a practical application and do not amount to significantly more than the abstract idea itself.
Claim 19 recites “wherein the virtual … machine learning model comprises a first large language model (LLM).” This recitation is at a high level of generality such that it amounts to no more than mere instructions to apply the exception using a generic computer element. Even in combination these additional elements do not integrate the abstract idea into a practical application and do not amount to significantly more than the abstract idea itself.
Claims 9 and 10 do not recite any additional elements, and accordingly, for the reasons provided above with respect to the independent claims, are not patent eligible.
Therefore, dependent claims 2-11, 13, 15-17, 19, 21 and 22 are not patent eligible.
Conclusion
The prior art made of record and not relied upon, as set forth in the accompanying Notice of References Cited (PTO-892), is considered pertinent to applicant's disclosure. Description of the cited references is provided above ("Subject Matter Distinguishable From Prior Art").
Any inquiry concerning this communication or earlier communications from the examiner should be directed to DOUGLAS W PINSKY whose telephone number is (571)272-4131. The examiner can normally be reached on 8:30 am - 5:30 pm ET.
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, Jessica Lemieux can be reached on 571-270-3445. 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.
/DOUGLAS W PINSKY/
Examiner, Art Unit 3626
1 Note: dependent claims 2 and 3 were not amended in the instant Response, and so instant dependent claims 2 and 3 still recite the content shown here.
2 "Contra proferentem is a rule of contract interpretation that states an ambiguous contract term should be construed against the drafter of the contract. … Contra proferentem exists to place the burden of ambiguity on the party most capable of mitigating that ambiguity – the person who wrote it." Legal Information Institute (LII) (Cornell Law School), definition of "contra proferentem," https://www.law.cornell.edu/wex/contra_proferentem.
Prosecution Insights