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 .
Specification
The disclosure is objected to because of the following informalities:
In paragraph 0014, line 7, “output by he service model” should read “output by the service model”.
In paragraph 0081, line 7, “output by he service model” should read “output by the service model”.
Appropriate correction is required.
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)(1) the claimed invention was patented, described in a printed publication, or in public use, on sale, or otherwise available to the public before the effective filing date of the claimed invention.
(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, 3 – 8, 10 – 15, and 17 – 20 are rejected under 35 U.S.C. 102(a)(1) as being anticipated by Nguyen et al. ("From Black Boxes to Conversations: Incorporating XAI in a Conversational Agent"), hereinafter Nguyen.
Regarding claim 1, Nguyen discloses a computer-implemented method for service execution, comprising:
acquiring a question statement of a user, wherein the question statement is determined based on a service result output by a service model (Abstract, lines 1-3, "The goal of Explainable AI (XAI) is to design methods to provide insights into the reasoning process of black-box models, such as deep neural networks, in order to explain them to humans."; Section 1, lines 19-23, "In this work, we develop methods to leverage a standard conversational agent architecture to conversational XAI. Building upon well-established conversational agent techniques allows us to focus on XAI-specific requirements, in order to cover a broad range of user questions, types of data, and types of models."; A user question reads on a question statement of a user, and providing insights into the reasoning process of black-box models, such as deep neural networks, reads on the question statement being determined based on a service result output by a service model.);
generating a query statement based on the question statement, wherein the query statement is used to query, based on, which the service model outputs the service result, user data of the user (Section 4.2, lines 1-16, "We preprocess a given user question to a standard format. We use a placeholder <feature> to substitute all feature names from the data set. Similarly, labels (classes) in user questions are replaced by the placeholder <class>. For example, on the Adult data set (See Footnote 1) (Fig. 1), the question How could I change only Occupation to get >50K? is transformed to How could I change only <feature> to get <class>?, in which, the Occupation is a feature and >50K is a class in the Adult data set. We assume that feature and class names in user questions match those in the data set (i.e., no typos or synonyms). We formulate the matching of a user question to a reference question as a multi-class classification problem with class labels corresponding to reference questions in the XAI question phrase bank (see Sect. 4.1). First, we generate sentence embeddings of the pre-processed user and reference questions with SimCSE [12] and RoBERTa-large [27]. We then train a feedforward network with 1 hidden layer and ReLU activation on the sentence embeddings to classify user questions into one of the reference questions in the question phrase bank. The output of this step is a reference question that reflects the intent of the user question."; Preprocessing a given user question to a standard format and classifying user questions into reference questions, where a reference question reflects the intent of the user question, reads on generating a query statement based on the question statement, wherein the query statement is used to query user data of the user.);
determining a query result based on the query statement (Section 5, lines 1-2, "After matching user input to its corresponding reference question, the next step is to obtain the relevant information to provide an answer."; Section 3.1, lines 3-6, "Questions in the data generation subcategory require information about the data or the data generation process. They can either be directly answered by querying data set statistics or accessing an accompanying data sheet for the data set."; Obtaining the relevant information to provide an answer reads on determining a query result based on the query statement.);
inputting the question statement into a recognition model that is pre-trained, to determine an interpretation strategy used for the question statement (Section 3, lines 11-17, "The general architecture of our conversational XAI agent is depicted in Fig. 2. The NLU component is responsible for identifying the user’s actual intent from a wide range of XAI utterances. To cope with this variety of utterances, we expand the XAI question bank [25] into an XAI question phrase bank. This phrase bank constitutes a training data set to identify user intents from a wide range of XAI utterances. We construct the extended question phrase bank from an initial set of XAI questions, paraphrase generation, and scoring."; Section 5, lines 9-14, "Specifically, we identified the questions that require an XAI method for extracting the answer information (highlighted rows), and not only require to access stored values, e.g. the size of the training data. Following the general definition of XAI by Arrieta et al. [4], we define an XAI method as a method that produces details or reasons to make the AI’s functioning clear or easy to understand."; Section 5.2, lines 56-57, "In summary, to integrate XAI into the dialogue policy component, we systematically map XAI questions to XAI methods."; A natural language understanding component identifying the user’s actual intent reads on inputting the question statement into a recognition model that is pre-trained, and mapping the question to an explainable artificial intelligence method reads on determining an interpretation strategy used for the question statement.);
determining, based on the query result through the interpretation strategy, an association relationship between at least part of the user data and the service result output by the service model (Section 5.2, lines 56-57, "In summary, to integrate XAI into the dialogue policy component, we systematically map XAI questions to XAI methods."; Section 6, lines 1-2, "XAI methods provide the core information to answer the corresponding questions, but they lack explanatory text in natural language for the end user."; An explainable artificial intelligence method reads on the interpretation strategy, and providing the core information to answer the corresponding questions reads on determining an association relationship between at least part of the user data and the service result output by the service model.);
and generating a reply statement for the question statement based on the association relationship, to execute a target service through the reply statement (Section 6, lines 1-11, "XAI methods provide the core information to answer the corresponding questions, but they lack explanatory text in natural language for the end user. Presenting just the raw information in form of a table or importance values alongside feature names to end users is not always adequate. Instead, additional context, such as what the values represent or how to interpret them, is desirable. To address this problem, we incorporate a template-based natural language generation component. For each question, we define text templates (partially with dataset-specific vocabulary) containing placeholders for the information obtained from XAI methods. We combine this information with – or convert it to – textual explanations depending on the type of information generated by the XAI method."; Providing explanatory text in natural language for the end user by converting the information obtained from explainable artificial intelligence methods to textual explanations reads on generating a reply statement for the question statement based on the association relationship.).
Regarding claim 3, Nguyen discloses the computer-implemented method as claimed in claim 1.
Nguyen further discloses:
wherein pre-training the recognition model comprises: acquiring each training sample, wherein the training sample is a simulation of the question statement determined by the user based on the service result output by the service model (Section 4.1, lines 1-5, "Training the NLU component of a conversational XAI agent requires conversational data about XAI. However, such data does not exist. Therefore, we introduce an XAI question phrase bank as a data set to train and evaluate the NLU component. The phrase bank represents a broad variety of utterances of possible user questions to XAI systems and is publicly available."; A question phrase bank training data set reads on the training sample.);
inputting the training sample into a to-be-trained recognition model, so that the to-be-trained recognition model extracts a feature of the training sample, and determines, based on extracted feature information, an interpretation strategy corresponding to the training sample (Section 4.2, lines 1-11, "We preprocess a given user question to a standard format. We use a placeholder <feature> to substitute all feature names from the data set. Similarly, labels (classes) in user questions are replaced by the placeholder <class>. For example, on the Adult data set (See Footnote 1) (Fig. 1), the question How could I change only Occupation to get >50K? is transformed to How could I change only <feature> to get <class>?, in which, the Occupation is a feature and >50K is a class in the Adult data set. We assume that feature and class names in user questions match those in the data set (i.e., no typos or synonyms). We formulate the matching of a user question to a reference question as a multi-class classification problem with class labels corresponding to reference questions in the XAI question phrase bank (see Sect. 4.1)."; Matching a user question to a reference question reads on determining an interpretation strategy corresponding to the training sample.);
and training the to-be-trained recognition model with an optimization target of minimizing a deviation between the interpretation strategy corresponding to the training sample and a tagged interpretation strategy corresponding to the training sample (Section 4.2, lines 11-20, "First, we generate sentence embeddings of the pre-processed user and reference questions with SimCSE [12] and RoBERTa-large [27]. We then train a feedforward network with 1 hidden layer and ReLU activation on the sentence embeddings to classify user questions into one of the reference questions in the question phrase bank. The output of this step is a reference question that reflects the intent of the user question. From the classifier’s output, we select the reference question with the highest probability. If the probability is lower than a predefined threshold θ (no match), we consider the question as an (yet) unknown variation (paraphrase) of a reference, save it for later, and ask the user for an alternative phrasing of the question."; Select the reference question with the highest probability reads on minimizing a deviation between the interpretation strategy corresponding to the training sample and a tagged interpretation strategy corresponding to the training sample.).
Regarding claim 4, Nguyen discloses the computer-implemented method as claimed in claim 1.
Nguyen further discloses:
wherein determining, based on the query result through the interpretation strategy, an association relationship between at least part of the user data and the service result output by the service model comprises: determining the at least part of the user data from the query result based on the interpretation strategy (Section 5.1, lines 3-6, "Questions in the data generation subcategory require information about the data or the data generation process. They can either be directly answered by querying data set statistics or accessing an accompanying data sheet for the data set [14] if available."; Section 5.2, lines 56-57, "In summary, to integrate XAI into the dialogue policy component, we systematically map XAI questions to XAI methods."; Section 6, lines 1-2, "XAI methods provide the core information to answer the corresponding questions, but they lack explanatory text in natural language for the end user."; Providing the core information to answer the corresponding questions reads on determining the at least part of the user data from the query result based on the interpretation strategy.).
Regarding claim 5, Nguyen discloses the computer-implemented method as claimed in claim 4.
Nguyen further discloses:
adjusting a user data value of the at least part of the user data, to obtain adjusted user data (Column 6, lines 18-25, "In case of counterfactual explanations, relations need to be extracted in addition to feature names and values. For the counterfactual question 53 (Why is this instance predicted P instead of Q?), we extract and compare the relation between the given instance with class P and a counterfactual instance with the target class Q. In particular, for numerical features, we identify the relation between two values, whether the first is greater or smaller than the second. In the case of categorical/textual features, we check if a value is different from the other."; A counterfactual instance for a numerical feature reads on adjusting a user data value.).
Regarding claim 6, Nguyen discloses the computer-implemented method as claimed in claim 5.
Nguyen further discloses:
inputting the adjusted user data into the service model, to obtain an adjusted service result (Section 5, lines 9-19, "Specifically, we identified the questions that require an XAI method for extracting the answer information (highlighted rows), and not only require to access stored values, e.g. the size of the training data. Following the general definition of XAI by Arrieta et al. [4], we define an XAI method as a method that produces details or reasons to make the AI’s functioning clear or easy to understand. That is, an XAI method must have access to a model’s internal reasoning or to a proxy that reveals this reasoning at least to some extent. For instance, the question Why is this instance predicted P instead of Q? requires a counterfactual explanation, identifying feature sets that – if changed – would change the model’s prediction from P to the specified counterfactual class Q."; Column 6, lines 18-25, "In case of counterfactual explanations, relations need to be extracted in addition to feature names and values. For the counterfactual question 53 (Why is this instance predicted P instead of Q?), we extract and compare the relation between the given instance with class P and a counterfactual instance with the target class Q. In particular, for numerical features, we identify the relation between two values, whether the first is greater or smaller than the second. In the case of categorical/textual features, we check if a value is different from the other."; Identifying feature sets that, if changed, change the model’s prediction reads on inputting the adjusted user data into the service model to obtain an adjusted service result.).
Regarding claim 7, Nguyen discloses the computer-implemented method as claimed in claim 6.
Nguyen further discloses:
determining a deviation between the adjusted service result and the service result output by the service model based on the query result (Section 5, lines 9-19, "Specifically, we identified the questions that require an XAI method for extracting the answer information (highlighted rows), and not only require to access stored values, e.g. the size of the training data. Following the general definition of XAI by Arrieta et al. [4], we define an XAI method as a method that produces details or reasons to make the AI’s functioning clear or easy to understand. That is, an XAI method must have access to a model’s internal reasoning or to a proxy that reveals this reasoning at least to some extent. For instance, the question Why is this instance predicted P instead of Q? requires a counterfactual explanation, identifying feature sets that – if changed – would change the model’s prediction from P to the specified counterfactual class Q."; Column 6, lines 18-25, "In case of counterfactual explanations, relations need to be extracted in addition to feature names and values. For the counterfactual question 53 (Why is this instance predicted P instead of Q?), we extract and compare the relation between the given instance with class P and a counterfactual instance with the target class Q. In particular, for numerical features, we identify the relation between two values, whether the first is greater or smaller than the second. In the case of categorical/textual features, we check if a value is different from the other."; A change the model’s prediction from P to the specified counterfactual class Q reads on a deviation between the adjusted service result and the service result output by the service model based on the query result.);
and determining, based on the deviation, the association relationship between the at least part of the user data and the service result output by the service model (Section 5, lines 9-19, "Specifically, we identified the questions that require an XAI method for extracting the answer information (highlighted rows), and not only require to access stored values, e.g. the size of the training data. Following the general definition of XAI by Arrieta et al. [4], we define an XAI method as a method that produces details or reasons to make the AI’s functioning clear or easy to understand. That is, an XAI method must have access to a model’s internal reasoning or to a proxy that reveals this reasoning at least to some extent. For instance, the question Why is this instance predicted P instead of Q? requires a counterfactual explanation, identifying feature sets that – if changed – would change the model’s prediction from P to the specified counterfactual class Q."; Column 6, lines 18-25, "In case of counterfactual explanations, relations need to be extracted in addition to feature names and values. For the counterfactual question 53 (Why is this instance predicted P instead of Q?), we extract and compare the relation between the given instance with class P and a counterfactual instance with the target class Q. In particular, for numerical features, we identify the relation between two values, whether the first is greater or smaller than the second. In the case of categorical/textual features, we check if a value is different from the other."; Identifying the relation between two values of numerical features that change the model’s prediction reads on determining the association relationship between the at least part of the user data and the service result output by the service model.).
Regarding claim 8, arguments analogous to claim 1 are applicable. In addition, Nguyen discloses a non-transitory, computer-readable medium storing one or more instructions executable by a computer system to perform one or more operations (Section 5.2, lines 10-13, "The paper should be accompanied by easy-to-use, publicly accessible source code in order to integrate the method into the conversational agent in a reasonable amount of time."; Software demonstrates a processor executing instructions from memory.), comprising the steps of claim 1.
Regarding claim 10, arguments analogous to claim 3 are applicable.
Regarding claim 11, arguments analogous to claim 4 are applicable.
Regarding claim 12, arguments analogous to claim 5 are applicable.
Regarding claim 13, arguments analogous to claim 6 are applicable.
Regarding claim 14, arguments analogous to claim 7 are applicable.
Regarding claim 15, arguments analogous to claim 1 are applicable. In addition, Nguyen discloses a computer-implemented system, comprising: one or more computers; and one or more computer memory devices interoperably coupled with the one or more computers and having tangible, non-transitory, machine-readable media storing one or more instructions that, when executed by the one or more computers, perform one or more operations (Section 5.2, lines 10-13, "The paper should be accompanied by easy-to-use, publicly accessible source code in order to integrate the method into the conversational agent in a reasonable amount of time."; Software demonstrates a processor executing instructions from memory.), comprising the steps of claim 1.
Regarding claim 17, arguments analogous to claim 3 are applicable.
Regarding claim 18, arguments analogous to claim 4 are applicable.
Regarding claim 19, arguments analogous to claim 5 are applicable.
Regarding claim 20, arguments analogous to claim 6 are applicable.
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, 9 and 16 are rejected under 35 U.S.C. 103 as being unpatentable over Nguyen in view of Chang et al. (US Patent No. 10,025,819), hereinafter Chang.
Regarding claim 2, Nguyen discloses the computer-implemented method as claimed in claim 1, but does not specifically disclose: wherein generating a query statement based on the question statement comprises: dividing the question statement, to obtain component words based on the division; performing part of speech tagging on the component words, to obtain a tagging result; determining, as a target component word based on the tagging result, a component word whose part of speech is a noun from the component words, and determining a corresponding query parameter in the query statement based on the target component word; and determining the query statement based on the corresponding query parameter.
Chang teaches:
dividing the question statement, to obtain component words based on the division (Column 3, lines 26-31, “The embodied techniques include receiving a natural language input, such as an utterance searching for information. This input can be processed to detect a sentence, identify words in the sentence, and tag the words with corresponding part of speech types (e.g., to indicate whether a word is a noun, verb, adjective, etc.)."; Receiving a natural language input such as an utterance searching for information and identifying words in the sentence reads on dividing the question statement to obtain component words based on the division.);
performing part of speech tagging on the component words, to obtain a tagging result (Column 3, lines 26-31, “The embodied techniques include receiving a natural language input, such as an utterance searching for information. This input can be processed to detect a sentence, identify words in the sentence, and tag the words with corresponding part of speech types (e.g., to indicate whether a word is a noun, verb, adjective, etc.)."; Tagging the words with corresponding part of speech types reads on performing part of speech tagging on the component words, to obtain a tagging result.);
determining, as a target component word based on the tagging result, a component word whose part of speech is a noun from the component words, and determining a corresponding query parameter in the query statement based on the target component word (Column 3, lines 38-50, "For example, the techniques can implement a machine learning classifier to predict, based on a pattern formed by the groupings, that one clause is associated with particular groupings while another clause is associated with other groupings. Once the prediction is available, words can then be added accordingly to the clauses to generate the query. For example, if the prediction indicates that a noun grouping (e.g., a grouping containing a noun tag corresponding to a noun) belongs to a select clause of an SQL query, the noun corresponding to that noun grouping can be added to the select clause. In comparison, if the prediction was for a where clause, the noun would be added to the where clause instead."; Indicating that a noun grouping belongs to a select clause of an SQL query and adding the noun corresponding to that noun grouping to the select clause reads on determining a component word whose part of speech is a noun from the component words, and determining a corresponding query parameter in the query statement based on the target component word.);
and determining the query statement based on the corresponding query parameter (Column 3, lines 38-50, "For example, the techniques can implement a machine learning classifier to predict, based on a pattern formed by the groupings, that one clause is associated with particular groupings while another clause is associated with other groupings. Once the prediction is available, words can then be added accordingly to the clauses to generate the query. For example, if the prediction indicates that a noun grouping (e.g., a grouping containing a noun tag corresponding to a noun) belongs to a select clause of an SQL query, the noun corresponding to that noun grouping can be added to the select clause. In comparison, if the prediction was for a where clause, the noun would be added to the where clause instead."; Adding words according to the clauses to generate the query reads on determining the query statement based on the corresponding query parameter.).
Chang is considered to be analogous to the claimed invention because it is in the same field of generating query statements. Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Nguyen to incorporate the teachings of Chang to receive a natural language input such as an utterance searching for information, identify words in the sentence, tag the words with corresponding part of speech types, indicate that a noun grouping belongs to a select clause of an SQL query, and add the noun corresponding to that noun grouping to the select clause to generate a query. Doing so would allow for generating a structured query to a dataset based on natural language search input (Chang; Column 2, lines 64-66).
Regarding claim 9, arguments analogous to claim 2 are applicable.
Regarding claim 16, arguments analogous to claim 2 are applicable.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure:
Arumugam et al. (US Patent No. 12,536,372) teaches a methods for receiving a query and target pair, providing a query-target prediction by processing the query entity and the target entity, generating a prompt by populating a prompt template with at least a portion of the query-target prediction, inputting the prompt into a large language model (LLM), and receiving, from the LLM, an explanation that is responsive to the prompt and that describes one or more reasons for the query-target prediction output by the entity matching model.
Satish et al. (US Patent No. 12,699,910) teaches a method for an explainability-augmented AI system configured to automatically identify, based on one or more of metadata associated with labels assigned to sample data and responses to AI-system-related questionnaires, one or more reasons that support the decisions made by an AI model in response to user queries.
Inavalli et al. (US Patent Application Publication No. 2024/0232294) teaches a method for combining structured and semi-structured data for explainable artificial intelligence (AI).
Rong et al. ("Towards Human-centered Explainable AI: A Survey of User Studies for Model Explanations") teaches an analysis of human-based explainable AI methods.
Zhang et al. ("May I Ask a Follow-up Question? Understanding the Benefits of Conversations in Neural Network Explainability") teaches a method for free-form conversations for explainable AI.
Slack et al. ("Explaining Machine Learning Models with Interactive Natural Language Conversations Using TalkToModel") teaches an interactive dialogue system that enables users to explain ML models through natural language conversations.
Chou et al. ("Counterfactuals and Causability in Explainable Artificial Intelligence: Theory, Algorithms, and Applications") teaches a review of counterfactuals and causability for explainable artificial intelligence.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to James Boggs whose telephone number is (571)272-2968. The examiner can normally be reached M-F 8:00 AM - 5:00 PM.
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, Daniel Washburn can be reached at (571)272-5551. 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.
/JAMES BOGGS/Examiner, Art Unit 2657
Prosecution Insights