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 .
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 1-9, 11-16 and 18-20 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.
The Supreme Court has long held that “[l]aws of nature, natural phenomena, and abstract ideas are not patentable.” Alice Corp. Pty. Ltd. v. CLS Bank Int’l, 134 S. Ct. 2347, 2354 (2014) (quoting Assoc. for Molecular Pathology v. Myriad Genetics, Inc., 133 S. Ct. 2107, 2116 (2013) (internal quotation marks omitted)). The “abstract ideas” category embodies the longstanding rule that an idea, by itself, is not patentable. Alice Corp., 134S. Ct. at 2355 (quoting Gottschalk v. Benson, 409 U.S. 63, 67 (1972).
In Alice, the Supreme Court sets forth an analytical “framework for distinguishing patents that claim laws of nature, natural phenomena, and abstract ideas [or mental processes ] from those that claim patent-eligible applications of those concepts.” Id. at 2355 (citing Mayo Collaborative Servs. v. Prometheus Labs., Inc., 132 S. Ct. 1289, 1296–97 (2012)). The first step in the analysis is to “determine whether the claims at issue are directed to one of those patent-ineligible concepts.” Id. If the claims are directed to a patent-ineligible concept, the second step in the analysis is to consider the elements of the claims “individually and ‘as an ordered combination’” to determine whether there are additional elements that “‘transform the nature of the claim’ into a patent-eligible application.” Id. (quoting Mayo, 132 S. Ct. at 1298, 1297). In other words, the second step is to “search for an ‘inventive concept’—i.e., an element or combination of elements that is ‘sufficient to ensure that the patent in practice amounts to significantly more than a patent upon the [ineligible concept] itself’”. Id. (brackets in original) (quoting Mayo, 132 S. Ct. at 1294). The prohibition against patenting an abstract idea “‘cannot be circumvented by attempting to limit the use of the formula to a particular technological environment’ or adding ‘insignificant post-solution activity.’” Bilski v. Kappos, 561 U.S. 593, 610–11 (2010) (citation omitted).
Step 1: This part of the eligibility analysis evaluates whether the claim falls within any statutory category. See MPEP 2106.03. Independent Claim 1 recites the method of intent matching of a determined intent of a user input and selectively providing responses based on whether the determined intent is matched with a first list of predefined intents or a second list of select intents. A process is a statutory category of invention. Independent Claim 12 recites a system comprising one or more processors and a memory configured to execute a method similar to Claim 1. A system or apparatus is a Statutory category of invention. Independent claim 18 recites a non-transitory computer-readable medium embodying program code that is executable by one or more processors to cause the one or more processors to perform steps similar to Claim 1. A non-transitory computer-readable medium is a statutory category. Dependent claims 2-11, 13-17 and 19-20 are dependent on claims 1, 12 and 18, respectively, and therefore recite their respective statutory classes.
Step 2A, Prong One: This part of the eligibility analysis evaluates whether the claim recites a judicial exception. As explained in MPEP 2106.04, subsection II, a claim “recites” a judicial exception when the judicial exception is “set forth” or “described” in the claim. In applying the framework set out in Alice, examiner found Applicant’s claims 1, 12 and 18 are directed to a patent-ineligible abstract concept of selectively providing responses to user inputs based on the result of intent matching. The steps of Applicant’s claims 1, 12 and 18 are an abstract concept that would fall under the judicial exception of mental processes. Specifically, the claims recite the step of “receiving a text input associated with a virtual interaction.” The recited receiving a text input involves nothing more than the transferring of data. Under broadest reasonable interpretation, the text may be received from a textual note written on a piece of paper. The recited “associated with a virtual interaction” represents nothing more than a generic computer that is in communication with another generic computer that enables the transferring of data in the virtual interaction. The recited virtual interaction represents an additional element that is nothing more than a generic step (see Step 2A, Prong Two). Therefore, this step is directed to a mental process. Furthermore, the step of “determining, using a first language model, an intent associated with the text input” recites a step that is directed to mental process. The claim does not place any limits to how the intent is determined. Under the broadest reasonable interpretation, determining an intent may be characterized by a human interpreting a received textual request. Although the claim recites a first language model to determine the intent, the first language model may be a database of information that a human may use to aid in the interpretation of a received textual request. Thus, the step is directed to a mental process. Further, the claim recites “receiving, by the first language model, a first list of predefined intents and a second list of select intents, wherein each predefined intent of the first list of predefined intents is associated with a respective predefined response from a list of predefined responses”. The recited limitation is directed to nothing more than the transfer of data. The claim does not place any limits on how the first and second lists are received. The claim simply recites that a list of “pre-defined” data is received. Thus, under the broadest reasonable interpretation the claim elements are directed to a mental process. Further, the claim recites the step of “comparing the determined intent to the first list of predefined intents and to the second list of select intents to determine a respective first match or a respective second match”. The claim does not place any limits on how the matches are determined, or what process is used to “match” the intents. Thus, under broadest reasonable interpretation, the recited comparison may be characterized by a human going through a list of pre-defined intents and using their own judgement to determine whether the human interpretation of the received textual request matches a pre-defined intent contained in the list of pre-defined intents. Therefore, the recited step is directed to a mental process. Further, the claim recites the step of “responsive to determining the respective first match, retrieving and outputting the respective predefined response.” The recited step is directed to nothing more than information retrieval, which is a step that a human is capable of performing in the mind or by using pen and paper. Thus, the step is directed to a mental process. Further, the claim recites the step of “responsive to determining the respective second match, providing the text input and a prompt to a second language model.” The recited step again recites nothing more than information retrieval. Under broadest reasonable interpretation, the recited step of “providing the text input and a prompt to a second language model” may be characterized by selectively transferring the request to a second human operator to handle the request when the second match is determined. The step does not place any limits on how the “prompt” is provided. Thus, the recited steps are directed to a mental process. Finally, the step of “generating and outputting a generative response to the text input using the second language model” fails to provide any limit on how “generative response” is generated. Under broadest reasonable interpretation, and following the same example as previously provided, the step may be characterized by the second human operator using their own judgement to interpret and provide a response to the textual request. Therefore, the step is directed to a mental process. The claims recite limitations that taken in combination, recite at least a series of mental processes.
Regarding dependent claim 2, the claim recites “wherein the text input comprises a plurality of words, each word comprising a plurality of characters, and wherein the method further comprises: parsing, using the first language model, the text input to determine a total number of characters associated with the text input; and responsive to determining that the total number of characters does not satisfy a threshold, generating and outputting an indication that the text input is non-compliant.” The recited steps are directed to mental processes because a human is capable of counting the mount of characters in a received textual input and making a decision of non-compliance upon determining that the amount of characters does not satisfy a threshold amount of characters.
Regarding dependent claim 3, the claim recites “determining a validity of the generative response by: computing, using the second language model, a confidence score associated with the generative response; evaluating the confidence score against one or more threshold confidence scores; responsive to determining the confidence score satisfies the one or more threshold confidence scores, tagging the generative response as valid and outputting the generative response; and responsive to determining the confidence score does not satisfy the one or more threshold confidence scores, generating a new generative response using the second language model by providing an updated prompt and the text input to the second language model.” The recited step of calculating a “confidence score” is not directed to a mental process, but it is directed to a mathematical algorithm used to calculate a numerical value representing the quality or accuracy of a response. Furthermore, the process of tagging and generating a new response are directed to a mental process that is recited as a result of the mathematical algorithm that governs the calculation of the confidence score. Therefore, the steps are directed to a combination of mathematical algorithm and mental steps.
Regarding dependent claim 4, the claim recites “wherein the virtual interaction is an online chat session and wherein the virtual interaction comprises real-time chat messages in the online chat session.” The recited step discloses a representation of the virtual interaction as a real-time online chat session, which constitutes an additional element of the judicial exception. Further analysis provided in Step 2A, Prong Two and Step 2B. Regarding dependent claim 5, the claim recites “responsive to generating and outputting the generative response, receiving a new text input from a user; extracting a second intent from the new text input; and generating and outputting, based on the second intent, a second response, wherein the second response is retrieved from the list of predefined responses or the second response is generated by the second language model.” The recites steps provide nothing more than a further iteration of the same steps of the independent claim upon receiving a new text input from a user. Thus, they constitute a mental process for the same reasons as stated above. Regarding dependent claim 6, the claim recites “retrieving, using the second language model, session data associated with the virtual interaction, wherein the generative response is generated at least in part based on the session data.” The recited steps provide no limit regarding what the session data is and how it is retrieved, thus the recitation constitutes nothing more than generic data retrieval, and is a mental process. Regarding dependent claim 7, the claim recites “wherein the session data is extracted from a publicly accessible webpage and is associated with the determined intent.” As stated before, the recited steps provide no limit regarding what the session data is and how it is retrieved, thus the recitation constitutes nothing more than generic data retrieval, and is also a mental process.
Regarding dependent claim 8, the claim recites “wherein the session data is associated with a user account associated with a user of the virtual interaction.” As stated before, the recited steps provide no limit regarding what the session data is and how it is retrieved, thus the recitation constitutes nothing more than generic data retrieval, and is also a mental process.
Regarding dependent claim 9, the claim recites “responsive to determining that the determined intent is not included in the first list of predefined intents or the second list of select intents, labeling the determined intent as an unrecognized intent; storing the unrecognized intent in a datastore comprising a plurality of unrecognized intents; clustering the plurality of unrecognized intents using a classification model to thereby generate one or more clusters, each respective cluster comprising a subset of unrecognized intents; responsive to the subset of unrecognized intents of a respective cluster satisfying a threshold, assigning a new select intent to the respective cluster; and adding the new select intent to the second list of select intents.” The recited steps are generally directed to clustering unrecognized intents using a classification model. The recited elements do not provide any limit to what kind of classification model is being employed and how the classification model is employed to cluster the intents. The clustering may be characterized by a human operator being supplied with general rules (classification model) on how to handle intents that do not appear on a list of predefined intents to organize them in a certain classification scheme. Further, the claim recites steps that are directed to updating a second list of intents once a respective cluster reaches a threshold, however, the recitation fails to place a limit on how the subset of unrecognized intents of a respective cluster is compared to the threshold. In other words, the claim fails to provide any specificity on what parameters associated with the “respective cluster” are compared to the threshold in order to satisfy the limitation. Further, the assigning and adding limitations may constitute a human operator manually updating a database based on the findings that were determined upon applying the general rules (classification model). Thus, the recited limitations are directed to mental processes.
As per dependent claim 10, the claim recites “wherein prior to processing the text input, the first language model and the second language model were generated by fine-tuning respective instances of a pre-trained language model, and wherein the method further comprises: fine-tuning the second language model based in part on the new select intent.” The recited fine-tuning steps provide an additional element that constitute a practical application. Further analysis provided in the Step 2A, Prong Two section below.
As per dependent claim 11, the claim recites “wherein determining the intent is performed by comparing the text input to a set of predetermined keywords.” The claimed step recites a simple word comparison of a text input to a set of predetermined keywords. A word comparison is an operation that a human is capable of performing in the mind and thus is a mental process.
Further dependent claims 13-17 and 19-20 recite similar language as dependent claims 2-11, and thus are analyzed in a similar manner.
Step 2A, Prong Two: This part of the eligibility analysis evaluates whether the claim as a whole integrates the recited judicial exception into a practical application of the exception. This evaluation is performed by (1) identifying whether there are any additional elements recited in the claim beyond the judicial exception, and (2) evaluating those additional elements individually and in combination to determine whether the claim as a whole integrates the exception into a practical application. See MPEP 2106.04(d).
As discussed above, the claims recite “…a virtual interaction” as an additional element beyond the judicial exception. The examiner has found, however, that the use of a “virtual interaction” step provides no further detail and is recited at such a high-level of generality that this limitation is merely a post-solution step. Therefore, this step is an insignificant extra-solution activity and does not integrate the judicial exception into a practical application. See MPEP 2106.05(g). Furthermore, independent claim 12 further recites “one or more processors; a memory coupled to the one or more processors, the memory including instructions that, when executed by the one or more processors, cause the one or more processors to…” as additional elements beyond the judicial exception. However, these additional elements do not amount to significantly more than the abstract idea because the additional elements constitute a generic computer environment. Alice, 134 S. Ct. at 2357. The Claims need meaningful limitations that go beyond generally linking the use of an abstract idea to a particular technological environment. Therefore, the steps are all abstract and the Claim as a whole is abstract. “[S]imply appending generic computer functionality to lend speed or efficiency to the performance of an otherwise abstract concept does not meaningfully limit claim scope for purposes of patent eligibility.” CLS Bank, 2013 U.S. App. LEXIS 9493, at *29 (citing Bancorp, 687 F.3d at 1278, and Dealertrack, Inc. v. Huber, 674 F.3d 1315, 1333-34 (Fed. Cir. 2012) (finding that the claimed computer-aided clearinghouse process is a patent-ineligible abstract idea)); SiRF Tech., Inc. v. Int'l Trade Comm'n, 601 F.3d 1319, 1333 (Fed. Cir. 2010) (“In order for the addition of a machine to impose a meaningful limit on the scope of a claim, it must play a significant part in permitting the claimed method to be performed, rather than function solely as an obvious mechanism for permitting a solution to be achieved more quickly, i.e., through the utilization of a computer for performing calculations.”). Additionally, dependent claim 4 recites the additional element of providing real-time chat messages through an online chat session. The additional element, however, provides no further detail and is recited at such a high-level of generality that the online chat session is merely a post solution step. The recited step may be characterized by a chat session that is conducted using a conventional computing network. Therefore, this step is an insignificant extra-solution activity and does not integrate the judicial exception into a practical application. See MPEP 2106.05(g).
Dependent claims 10 and 17 recite “…fine-tuning the second language model based in part on the new select intent.” The recited element constitutes a practical application because fine-tuning the second language model using the findings of the clustering operations provides an improvement in the particular technological environment. Thus, the recited language of claims 10 and 17 are eligible and no further analysis is required.
Step 2B: This part of the eligibility analysis evaluates whether the claim as a whole amounts to significantly more than the recited exception, i.e., whether any additional element, or combination of additional elements, adds an inventive concept to the claim. See MPEP 2106.05.
At step 2A, prong two, the additional elements of a virtual interaction as an online chat session and the “one or more processors…” and “memory…” were found to be insignificant extra-solution activity and a generic computer environment. At Step 2B, the re-evaluation of the insignificant extra-solution activity consideration takes into account whether or not the extra-solution activity is well understood, routine, and conventional in the field. See MPEP 2106.05(g). Here, the step of providing the virtual interaction as an online chat session is an extra solution activity that is well-understood, routine and conventional, because an online chat session may be provided through a conventional computing network. Therefore, this limitation remains insignificant extra-solution activity even upon reconsideration and does not amount to significantly more. Even when considered in combination, these additional elements represent mere instructions to apply an exception and insignificant extra-solution activity, and therefore do not provide an inventive concept.
In conclusion, Examiner notes that none of recited steps in Applicant's claims 1-9, 11-16 and 18-20 refer to a specific machine by reciting structural limitations of any apparatus or to any specific operations that would cause a machine to be the mechanism to perform these steps. Although the claims may be processed by a computing system having a processor, the computing system is merely a general purpose computing system. Therefore, all of the claims 1-9, 11-16 and 18-20 are abstract.
Claim Rejections - 35 USC § 102
The following is a quotation of the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(2) the claimed invention was described in a patent issued under section 151, or in an application for patent published or deemed published under section 122(b), in which the patent or application, as the case may be, names another inventor and was effectively filed before the effective filing date of the claimed invention.
Claims 1, 4-8, 12, 16 and 18-20 are rejected under 35 U.S.C. 102(a)(2) as being anticipated by Moise (US PG Pub 20250384880).
As per claims 1, 12 and 18, Moise discloses: A method and a system comprising: one or more processors (Moise; Fig. 6, item 602; p. 0062-0064); a memory coupled to the one or more processors, the memory including instructions (Moise; Fig. 6, item 612; p. 0064 - Processor(s) 602 are generally configured to retrieve and execute instructions stored in one or more memories, including local memories like the computer-readable medium 612, as well as remote memories and data stores…; see also p. 0068 - Computer-readable medium 612 may be a volatile memory), and
a non-transitory computer-readable medium embodying program code that is executable by the one or more processors (Moise; Fig. 6, item 612; p. 0064 - Processor(s) 602 are generally configured to retrieve and execute instructions stored in one or more memories, including local memories like the computer-readable medium 612, as well as remote memories and data stores…; see also p. 0068 - Computer-readable medium 612 may be a volatile memory) to cause the one or more processors to: receive a text input associated with a virtual interaction (Moise; Fig. 2, item 204; p. 0030 - the UI 302 receives the user utterance, at 204, as a text message typed by the customer into a text field of a chat module provided on a website, or in a product, for example); determine, using a first language model, an intent associated with the text input (Moise; Fig. 3, item 310; p. 0033 - The smart dispatcher 310 includes a classifier model (first language model) that identifies the intent of the customer (e.g., user intent) by processing the user utterance and context provided in the user request; see also p. 0054-0056); receive, by the first language model, a first list of predefined intents and a second list of select intents (Moise; p. 0034 - A fulfillable user intent is an intent that has a matching intent in a list of previously stored intents (e.g., master intent list). The previously stored intents may be arranged in a lookup table, database, or similar data structures. Each of the previously stored intents may be associated with a unique intent ID; see also p. 0045; see also p. 0054-0056 - …At 506, the method may, in accordance with certain aspects of the present disclosure, compare the user intent against a master intent list (e.g., list of previously stored intents)…; see also p. 0036 - Each of the intent IDs in the configuration includes one or more human-curated responses from the database of human-curated responses 324. Additionally the configuration may also include flags that define whether the configuration is one that should be handled only by… the GenAI module 330… the flag that defines whether the configuration is one that should be handled only by the GenAI, represents the recited “second list of select intents”, because the items that include this flag represent a subset of the master intent list. As disclosed in the specification of the instant application in p. 0025, the “second list of select intents” is a sub-list of the list of predefined intents), wherein each predefined intent of the first list of predefined intents is associated with a respective predefined response from a list of predefined responses (Moise; p. 0036 - Each configuration may be a lookup table or database of related intent IDs corresponding to the domain of the configuration. Each of the intent IDs in the configuration includes one or more human-curated responses from the database of human-curated responses 324; see also p. 0054-0056 - …The master intent list may include a list of intents for which a set of human-curated responses are available through the first AI model...); compare the determined intent to the first list of predefined intents and to the second list of select intents to determine a respective first match or a respective second match (Moise; p. 0034 - A fulfillable user intent is an intent that has a matching intent in a list of previously stored intents (e.g., master intent list). The previously stored intents may be arranged in a lookup table, database, or similar data structures. Each of the previously stored intents may be associated with a unique intent ID; see also p. 0045; see also p. 0054-0056 - …At 506, the method may, in accordance with certain aspects of the present disclosure, compare the user intent against a master intent list (e.g., list of previously stored intents)…; see also p. 0036 - Each of the intent IDs in the configuration includes one or more human-curated responses from the database of human-curated responses 324. Additionally the configuration may also include flags that define whether the configuration is one that should be handled only by… the GenAI module 330… the flag that defines whether the configuration is one that should be handled only by the GenAI, represents the recited “second list of select intents”, because the items that include this flag represent a subset of the master intent list. As disclosed in the specification of the instant application in p. 0025, the “second list of select intents” is a sub-list of the list of predefined intents); responsive to determining the respective first match, retrieve and output the respective predefined response (Maoise; p. 0039-0041 - …if at least one configuration has an intent ID matching the user intent ID at 230, then the smart dispatcher 310 transmits the user request to a digital assistant 312. At 232, the digital assistant retrieves a human-curated response associated with the configuration and user intent ID from the database of human-curated responses 324 and transmits the response to the UI 302 and provided as an answer to the customer at 236…); responsive to determining the respective second match, provide the text input and a prompt to a second language model (Moise; p. 0035 - At 216, if the user intent is determined to not be fulfillable by the conversational experiences platform 320, the routing model signals the smart dispatcher to forward the user utterance to the GenAI module 330. The smart dispatcher 310, upon receiving the signal from the routing model 314, at 218 forwards the user request to the GenAI module 330, as shown in FIG. 3B); and generate and output a generative response to the text input using the second language model (Moise; p. 0035 - At 220, the GenAI module 330, processes the user request and generates a response. The generated response is transmitted to the UI 302 and provided as an answer to the customer at 236).
As per claim 4, Moise discloses:
The method of claim 1, wherein the virtual interaction is an online chat session and wherein the virtual interaction comprises real-time chat messages in the online chat session (Moise; p. 0030 - the UI 302 receives the user utterance, at 204, as a text message typed by the customer into a text field of a chat module provided on a website (online)…; see also p. 0042 - the GenAI module 330… generates each response in real time).
As per claims 5 and 20, Moise discloses: The method and non-transitory computer-readable medium of claims 1 and 18, further comprising program code that is executable by the one or more processors to cause the one or more processors to: responsive to generating and outputting the generative response, receiving a new text input from a user (Moise; p. 0042 - …The smart dispatcher 310 may refer to the user conversation list each time a new utterance is received from the customer…; see also p. 0021); extracting a second intent from the new text input (Moise; p. 0042 - …As long as the current utterance has the same user intent ID, the digital assistant 312 may use one of the follow-up responses as an answer. However, if the user issues an utterance that shifts the conversation to a new user intent—thus, causing the user intent ID to change—the smart dispatcher 310 processes the new user request, as previously described, by returning to 210 and proceeding accordingly…; see also p. 0021); and generating and outputting, based on the second intent, a second response, wherein the second response is retrieved from the list of predefined responses or the second response is generated by the second language model (Moise; p. 0042 - …Each time the user intent ID changes the user conversation list is cleared and populated with new follow-up responses associated with the new user intent ID; see also p. 0021).
As per claim 6, Moise discloses: The method of claim 1, further comprising: retrieving, using the second language model, session data associated with the virtual interaction, wherein the generative response is generated at least in part based on the session data (Moise; p. 0021 - …the smart dispatcher maintains a conversation list (session data) that includes a list of follow-up intents that reflect probably subsequent related intents that may arise during the conversation with the user. The conversation list is maintained and updated throughout the conversation while responses are provided by the NLU operating mode. Moreover the NLU operating mode applies rules and templates to the human-curated responses to personalize responses based on user information).
As per claim 7, Moise discloses: The method of claim 6, wherein the session data is extracted from a publicly accessible webpage and is associated with the determined intent (Moise; p. 0028 - The GenAI module 108, may include a GenAI model, such as a pre-trained LLM model, that is trained on a set of relevant datasets. For example, the LLM may be trained using manuals and documents relating to the functions of a product or service (publicly accessible datasets). In cases where the product is a financial product, such as tax preparation software, and the like, the LLM model may be trained on tax and other finance related data).
As per claim 8, Moise discloses: The method of claim 6, wherein the session data is associated with a user account associated with a user of the virtual interaction (Moise; p. 0071 - The request generating component 614 may also retrieve customer information 636 and experience information 632, which may be provided through the user interface, or stored in a customer database, for example. The request generating component 614 combines the user utterance, the customer information 636 and experience information 632 into a user request).
As per claim 16, the claim is directed to a system that contains subject matter similar to the combination of claims 6 and 7. Therefore, the claim is rejected similarly, in view of Moise.
As per claim 19, the claim is directed to a non-transitory computer-readable medium that contains subject matter similar to the combination of claims 6 and 7. Therefore, the claim is rejected similarly, in view of Moise.
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 2 and 13 are rejected under 35 U.S.C. 103 as being unpatentable over Moise in view of Karpman et al. (US Patent 11,995,803; hereinafter “Karpman”).
As per claims 2 and 13, Moise discloses: The method and system of claims 1 and 12, wherein the text input comprises a plurality of words, each word comprising a plurality of characters (Moise; Fig. 2, item 204; p. 0030 - the UI 302 receives the user utterance, at 204, as a text message typed by the customer into a text field of a chat module provided on a website, or in a product, for example; see also p. 0031 - …The user request may be formatted as a structured text elements, such as JSON, XML, or in other appropriate text-based format…), and wherein the method further comprises: parsing, using the first language model, the text input to determine a total number of characters associated with the text input; and responsive to determining that the total number of characters does not satisfy a threshold, generating and outputting an indication that the text input is non-compliant. Moise, however, fails to disclose parsing, using the first language model, the text input to determine a total number of characters associated with the text input; and responsive to determining that the total number of characters does not satisfy a threshold, generating and outputting an indication that the text input is non-compliant. Karpman does teach parsing, using the first language model, the text input to determine a total number of characters associated with the text input; and responsive to determining that the total number of characters does not satisfy a threshold, generating and outputting an indication that the text input is non-compliant (Karpman; Col. 22, lines 20-40 - In implementations, acceptable text prompts may be limited to a fixed number of characters, words, and/or tokens in order to reduce the amount of time needed for image generation and/or reduce serving latency. In these implementations, the software application layer 124 can calculate a word, character, or token count from the text prompt (e.g., text entered or input by the user into the interactive text field after any modifications or edits made by the software application layer 124 as described above); compare the word, character, or token count to a preset limit (threshold); and in response to determining that the word, character, or token count meets (e.g., exceeds) the preset limit (does not satisfy a threshold), decline to send the image generation task to the communication and cause the generation interface to display an error message near the interactive text field (outputting an indication that the text input is non-compliant)…). Therefore, it would have been obvious to one of ordinary skill in the art to modify the method and system of Moise to include parsing, using the first language model, the text input to determine a total number of characters associated with the text input; and responsive to determining that the total number of characters does not satisfy a threshold, generating and outputting an indication that the text input is non-compliant, as taught by Karpman, in order to reduce the amount of time needed for image generation and/or reduce serving latency (Karpman; Col. 22, lines 20-40).
Claims 3 and 14 are rejected under 35 U.S.C. 103 as being unpatentable over Moise in view of Tiwari (US PG Pub 20260093931).
As per claims 3 and 14, Moise discloses: The method and system of claims 1 and 12, upon which claims 3 and 14 depend.
Moise, however, fails to disclose the method further comprising: determining a validity of the generative response by: computing, using the second language model, a confidence score associated with the generative response; evaluating the confidence score against one or more threshold confidence scores; responsive to determining the confidence score satisfies the one or more threshold confidence scores, tagging the generative response as valid and outputting the generative response; and responsive to determining the confidence score does not satisfy the one or more threshold confidence scores, generating a new generative response using the second language model by providing an updated prompt and the text input to the second language model. Tiwari does teach determining a validity of the generative response by: computing, using the second language model, a confidence score associated with the generative response (Tiwari; p. 0062 - …The hallucination detection server 302, upon obtaining the query and the response, may evaluate the reliability of the response. The hallucination detection server 302 may apply multiple hallucination evaluation techniques to the query and/or the response to generate multiple hallucination scores (e.g., using modules 308a-308g). The hallucination detection server 302 may then aggregate or combine the plurality of hallucination scores in any suitable manner (e.g., using aggregation module 308h) to generate a confidence score, which indicates the reliability of the response…); evaluating the confidence score against one or more threshold confidence scores (Tiwari; p. 0041 - …The confidence metric may be evaluated and compared to a certain threshold, such as a predetermined or user-determined threshold for the relevancy of the generated answer data…; see also p. 0158 - …the aggregation module 308h may categorize the confidence score by comparing the confidence score to one or more threshold…); responsive to determining the confidence score satisfies the one or more threshold confidence scores, tagging the generative response as valid and outputting the generative response (Tiwari; p. 0158 - …the aggregation module 308h may categorize the confidence score by comparing the confidence score to one or more thresholds. For example, if the confidence score is above a first threshold (0.85), the aggregation module 308h may categorize the confidence score as high (tagging)…; see also p. 0160 where the categorization (or tagging) of the response with a confidence score is used to adjust the language model using a reward system that rewards high confidence responses and punish low confidence responses using a reinforcement learning technique); and responsive to determining the confidence score does not satisfy the one or more threshold confidence scores, generating a new generative response using the second language model by providing an updated prompt and the text input to the second language model (Tiwari; p. 0159 - …if the aggregation module 308h determines that the confidence score is below a threshold score (e.g., the second threshold), the aggregation module 308h may provide the query to the language model of the language model server 202 to reevaluate the query (updated prompt and the text input or query) and generate a new response…; see also p. 0160 where the categorization (or tagging) of the response with a confidence score is used to adjust the language model using a reward system that rewards high confidence responses and punish low confidence responses using a reinforcement learning technique. Thus, when a lower confidence score response is categorized (tagged) for an input query, the language model is adjusted through the reinforcement learning technique, which would correspond to the prompt being “updated”).
Therefore, it would have been obvious to one of ordinary skill in the art to modify the method and system of Moise to include determining a validity of the generative response by: computing, using the second language model, a confidence score associated with the generative response; evaluating the confidence score against one or more threshold confidence scores; responsive to determining the confidence score satisfies the one or more threshold confidence scores, tagging the generative response as valid and outputting the generative response; and responsive to determining the confidence score does not satisfy the one or more threshold confidence scores, generating a new generative response using the second language model by providing an updated prompt and the text input to the second language model, as taught by Tiwari, in order to provide a comprehensive hallucination evaluation system that utilizes both preventative and detective measures to evaluate hallucinations. In this manner, the comprehensive hallucination evaluation system determines the likelihood that the response generated by the language model is a hallucination based on the confidence score. By using multiple hallucination evaluation techniques and combining them, the comprehensive hallucination evaluation system improves the accuracy and reliability of detecting hallucinations in a response. Accordingly, the comprehensive hallucination evaluation provides a detailed and precise evaluation of the response, enabling users to gauge the trustworthiness of the information (Fieldman; p. 0041).
Claims 9-11 15 and 17 are rejected under 35 U.S.C. 103 as being unpatentable over Moise in view of Hao et al. (US Patent 11,907,673; hereinafter “Hao”).
As per claim 9, Moise discloses: The method of claim 1, upon which claim 9 depends. Moise, however fails to disclose responsive to determining that the determined intent is not included in the first list of predefined intents or the second list of select intents, labeling the determined intent as an unrecognized intent; storing the unrecognized intent in a datastore comprising a plurality of unrecognized intents; clustering the plurality of unrecognized intents using a classification model to thereby generate one or more clusters, each respective cluster comprising a subset of unrecognized intents; responsive to the subset of unrecognized intents of a respective cluster satisfying a threshold, assigning a new select intent to the respective cluster; and adding the new select intent to the second list of select intents. Hao does teach responsive to determining that the determined intent is not included in the first list of predefined intents or the second list of select intents, labeling the determined intent as an unrecognized intent (Hao; Col. 7, lines 1-18 - …module 108a determines whether each user message has been associated with a user intent by chatbot application 103… For user messages that are not associated with an intent, module 108a classifies these messages as incomprehensible); storing the unrecognized intent in a datastore comprising a plurality of unrecognized intents (Hao; Col. 5, lines 40-42 - Chat messages exchanged between client computing device 102 and chatbot application 103 are stored in chat messages database 112; see also Col. 6, lines 22-29 - …Chat message database 112 is… configured to receive, generate, and store specific segments of data relating to the process of enhancing chatbot recognition of user intent through graph analysis…; see also Col. 10, lines 43-47 - module 108e can store the learned user intents in database 112 for retrieval by chatbot application 103 for use in one or more software modules of chatbot 103 for natural language processing (NLP), chat workflow processing, intent determination, and the like); clustering the plurality of unrecognized intents using a classification model to thereby generate one or more clusters, each respective cluster comprising a subset of unrecognized intents (Hao; Fig. 2, item 204; Col. 8, lines 32-40 - Clustering module 108b also arranges (step 204) the incomprehensible user messages into a second plurality of clusters. For the incomprehensible user messages, module 108b assigns the messages to a predetermined number of clusters (e.g., 200 clusters) and the clusters of the incomprehensible user messages are used as input to the classification model 110 (after training on the comprehensible training dataset described above) for prediction of user intent); responsive to the subset of unrecognized intents of a respective cluster satisfying a threshold, assigning a new select intent to the respective cluster (Hao; Col. 10, lines 16-35 - Next, model training and execution module 108e uses the first graph kernel matrix k from the training data to train (step 210) classification model 110. In some embodiments, classification model 110 is a Kernel SVM classifier. Module 108e trains classification model 110 using the first graph kernel matrix (that is generated from comprehensible user messages) to be able to predict user intents for a subsequent input graph kernel matrix that is based upon incomprehensible user messages. After classification model 110 is trained on the first graph kernel matrix k, model training and execution module 108e executes (step 212) the trained classification model using the second graph kernel matrix y as input to generate a predicted intent identifier for one or more of the incomprehensible user messages… the teaching of Hao provides for assigning the predicted intent identifier to the one or more incomprehensible user messages based on the comprehensible user messages… After execution of the trained model 110 on the graph kernel matrix y associated with the incomprehensible user messages, model training and execution module 108e determines one or more existing user intents that apply to one or more of the incomprehensible user messages—thereby converting the incomprehensible messages to comprehensible messages); and adding the new select intent to the second list of select intents (Hao; Col. 10, lines 30-47 - After execution of the trained model 110 on the graph kernel matrix y associated with the incomprehensible user messages, model training and execution module 108e determines one or more existing user intents that apply to one or more of the incomprehensible user messages—thereby converting the incomprehensible messages to comprehensible messages). Therefore, it would have been obvious to one of ordinary skill in the art to modify the method, system and non-transitory computer-readable medium of Moise to include responsive to determining that the determined intent is not included in the first list of predefined intents or the second list of select intents, labeling the determined intent as an unrecognized intent; storing the unrecognized intent in a datastore comprising a plurality of unrecognized intents; clustering the plurality of unrecognized intents using a classification model to thereby generate one or more clusters, each respective cluster comprising a subset of unrecognized intents; responsive to the subset of unrecognized intents of a respective cluster satisfying a threshold, assigning a new select intent to the respective cluster; and adding the new select intent to the second list of select intents, as taught by Hao, in order to improve the chatbot performance by re-training the underlying intent recognition model so that the model better understands the intent behind the requests/messages originating from end users (Hao; Col. 1, lines 56-60).
As per claim 10, Moise in view of Hao discloses:
The method of claim 9, wherein prior to processing the text input, the first language model and the second language model were generated by fine-tuning respective instances of a pre-trained language model (Moise; p. 0070 - The second AI component 626 generates personalized responses to user utterances based on pre-training and fine tuning performed on the second AI component 626). And further, Hao teaches wherein the method further comprises: fine-tuning the second language model based in part on the new select intent (Hao; Col. 7, lines 31-49 - clustering module 108b utilizes a fine-tuning model approach which introduces minimal task-specific parameters and can be trained on the downstream tasks by fine tuning the pre-trained parameters. In some embodiments, module 108b converts the user messages into vectors that can be interpreted by a Bidirectional Encoder Representations from Transformers (BERT) implementation (as described in J. Devlin et al., “BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding,” arXiv:1810.04805 [cs.CL] 24 May 2019, available at arxiv.org/pdf/1810.04805, which is incorporated herein by reference). BERT is an NLP model proposed by Google in 2018, which learns a word's different semantic information based on its left and right context and obtains token level and sentence level representation by using Masked Language Model and next sentence prediction. Clustering module 108b uses BERT to generate sentence-level representation (i.e., a vector ν for each sentence, or user message)).
Therefore, it would have been obvious to one of ordinary skill in the art to modify the method, system and non-transitory computer-readable medium of Moise to include wherein the method further comprises: fine-tuning the second language model based in part on the new select intent, as taught by Hao, in order to improve the chatbot performance by re-training the underlying intent recognition model so that the model better understands the intent behind the requests/messages originating from end users (Hao; Col. 1, lines 56-60).
As per claim 11, Moise discloses:
The method of claim 1, upon which claim 11 depends. Moise, however, fails to disclose wherein determining the intent is performed by comparing the text input to a set of predetermined keywords. Hao does teach wherein determining the intent is performed by comparing the text input to a set of predetermined keywords (Hao; Col. 3, lines 28-30 - each sub-graph data structure is based upon a plurality of keywords from the user messages associated with the corresponding cluster; see also Col. 13, lines 9-26 - …module 108i is configured with one or more predefined rules that operate to match certain words or phrases between the incomprehensible user message and the chatlog user message(s) to determine one or more chatlog user messages that are the most similar to the incomprehensible user message…).
Therefore, it would have been obvious to one of ordinary skill in the art to modify the method, system and non-transitory computer-readable medium of Moise to include wherein determining the intent is performed by comparing the text input to a set of predetermined keywords, as taught by Hao, in order to improve the chatbot performance by re-training the underlying intent recognition model so that the model better understands the intent behind the requests/messages originating from end users (Hao; Col. 1, lines 56-60).
As per claim 15, the claim is directed to a system that contains subject matter similar to the combination of claims 4 and 11. Therefore, the claim is rejected similarly, in view of Moise and Hao. As per claim 17, the claim is directed to a system that contains subject matter similar to the combination of claims 9 and 10. Therefore, the claim is rejected similarly, in view of Moise and Hao.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure. The prior art made of record and not relied upon includes:
Donaldson (US PG Pub 20260111676) discloses an automated response system for generating a response to a user input, the system comprising a first computer device, the first computer device comprising a processor for: receiving an input from a user; determining an intent of the input; based on the intent, determining a response from a pre-generated set of responses; and outputting the response to the user (Donaldson; Abstract).
Rajakaruna (US PG Pub 20230306962) discloses a system and a method to create a conversational artificial intelligence is disclosed. The presented system is uniquely designed to drive more human-like but yet robust conversations via text or voice. In at least one embodiment, the disclosed system is implemented on at least one computing device that can respond to at least one communicating entity which is hereinafter referred to as the user. The system can be configured to drive a conversation in any knowledge domain. The system disclosed herein, uses natural language processing techniques, natural language synthesis techniques and a novel strategy to respond to user inputs. The novel strategy may include a fundamental model of human conversation, an algorithm to correlate user inputs to the aforesaid model, an algorithm to handle questions from the user, an algorithm to avoid undesired inputs from the user and finally an algorithm to predict the next response generated by the system (Rajakaruna; Abstract).
Any inquiry concerning this communication or earlier communications from the examiner should be directed to Rodrigo A Chavez whose telephone number is (571)270-0139. The examiner can normally be reached Monday - Friday 9-6 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, Richemond Dorvil can be reached at 5712727602. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/RODRIGO A CHAVEZ/Examiner, Art Unit 2658
/RICHEMOND DORVIL/Supervisory Patent Examiner, Art Unit 2658
Prosecution Insights