CTNF 18/458,135 CTNF 97128 DETAILED ACTION Notice of Pre-AIA or AIA Status 07-03-aia AIA 15-10-aia 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 07-04-01 AIA 07-04 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. Step 1: According to the first part of the analysis, in the instant case, claims 1-9, and 17-19 are directed to an apparatus claim comprising one or more processors and memory. Thus, each of the claims falls within one of the four statutory categories (i.e. process, machine, manufacture, or composition of matter). Claims 1-9 and 1 7-19 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more. In regard to claim 1 : Step 2A Prong 1: “ derive an evaluation value in the machine learning model for each document data included in the document data group” is a mental step of data manipulation. Additional Elements Step 2A Prong 2: “ An information processing apparatus comprising: at least one processor, wherein the processor is configured to: “does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)). “ for a machine learning model that uses a document data group including a plurality of document data as input and outputs output data” does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)). Step 2B: “ An information processing apparatus comprising: at least one processor, wherein the processor is configured to: “does not amount significantly more than the judicial exception in the claim. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)). “ for a machine learning model that uses a document data group including a plurality of document data as input and outputs output data” does not amount significantly more than the judicial exception in the claim. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)). In regard to claim 2 : Step 2A Prong 1: “from the document data group or specification of a display order of a document according to the document data based on the derived evaluation value” is a mental step of data comparison. Additional Elements Step 2A Prong 2: “ wherein the processor is configured to perform at least one of specification of the document data, which is a display target” does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)). Step 2B: “ wherein the processor is configured to perform at least one of specification of the document data, which is a display target” does not amount significantly more than the judicial exception in the claim. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)). In regard to claim 3 : Step 2A Prong 1: “ and derive the evaluation value for each document data based on the document unit output data” is a mental step of data manipulation. Additional Elements Step 2A Prong 2: “ wherein the processor is configured to: use each document data as input of the machine learning model to acquire document unit output data which is output for each document data” does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)). Step 2B: “ wherein the processor is configured to: use each document data as input of the machine learning model to acquire document unit output data which is output for each document data” does not amount significantly more than the judicial exception in the claim. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)). In regard to claim 4 : Step 2A Prong 1: “ wherein the evaluation value has a correlation with the document unit output data” is a mental step of data comparison. Step 2A Prong 2: no additional elements Step 2B: no additional elements In regard to claim 5 : Step 2A Prong 1: “ normalize each document data included in the document data group” is a mental step of data manipulation. “ and derive the evaluation value for each normalized document data” is a mental step of data manipulation. Additional Elements Step 2A Prong 2: “ wherein the processor is configured to: “does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)). Step 2B: “ wherein the processor is configured to: “does not amount significantly more than the judicial exception in the claim. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)). In regard to claim 6 : “ extract a plurality of word data from each document data included in the document data group” is a mental step of data manipulation. “ the evaluation value in the machine learning model as a word unit evaluation value for each word data” is a mental step of data manipulation. “ and derive the evaluation value according to a statistical value of the word unit evaluation value of the word data included in the document data for each document data” is a math function using pen and paper. Additional Elements Step 2A Prong 2: “ wherein the processor is configured to: “does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)). Step 2B: “ wherein the processor is configured to: “does not amount significantly more than the judicial exception in the claim. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)). In regard to claim 7 : Step 2A Prong 1: “ the document data having a greatest first evaluation value, which is derived for each document data, is used as first document data, and each of the plurality of document data other than the first document data included in the document data group is used as second document data” is a mental step of data identification. Additional Elements Step 2A Prong 2: “ and the processor is configured to use each combination data in which the first document data and the second document data are combined as input of the machine learning model to derive a second evaluation value from output data which is output for each combination data” does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)). Step 2B: “ and the processor is configured to use each combination data in which the first document data and the second document data are combined as input of the machine learning model to derive a second evaluation value from output data which is output for each combination data” does not amount significantly more than the judicial exception in the claim. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)). In regard to claim 8: Step 2A Prong 1: “ give a first display priority to the first document data” is a mental step of data manipulation. “ and give a second display priority, which is lower than the first display priority, to the second document data based on the second evaluation value” is a mental step of data manipulation. Additional Elements Step 2A Prong 2: “ wherein the processor is configured to: “does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)). Step 2B: “ wherein the processor is configured to: “does not amount significantly more than the judicial exception in the claim. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)). In regard to claim 9 : Step 2A Prong 1: “ extract a plurality of word data from each document data included in the document data Group” is is a mental step of data identification. “ derive the evaluation value in the machine learning model as a word unit evaluation value for each word data” is a math function using pen and paper. “ derive a first statistical value of the word unit evaluation value of the word data included in the document data for each document data to give a first evaluation value to first evaluation value document data which is the document data having a greatest first statistical value” is a mental step of data manipulation. “ derive, for a plurality of combination data in which the first evaluation value document data, and each of the plurality of document data other than the first evaluation value document data included in the document data group are combined, a second statistical value of the word unit evaluation value of the word data included in the combination data for each combination data to give a second evaluation value, which is lower than the first evaluation value, to second evaluation value document data which is the document data having a greatest second statistical value” is a mental step of data manipulation. “ and set, in derivation of the second statistical value, the word unit evaluation value of the word data included in the first evaluation value document data among the word data included in the document data combined with the first evaluation value document data to be relatively lower than the word unit evaluation value of the word data which is not included in the first evaluation value document data” is a mental step of data manipulation. Additional Elements Step 2A Prong 2: “ wherein the processor is configured to: “ does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)). Step 2B: “ wherein the processor is configured to: “ does not amount significantly more than the judicial exception in the claim. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)). In regard to claim 17 : Step 2A Prong 1: “ derive an evaluation value in the machine learning model for each document data included in the document data group” is a mental step of data manipulation. Additional Elements Step 2A Prong 2: “ An information processing apparatus comprising: at least one processor, wherein the processor is configured to: “does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)). “ for a machine learning model that uses a document data group including a plurality of document data as input and outputs output data” does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)). “ wherein the machine learning model is a machine learning model trained by a learning apparatus of the machine learning model that uses the document data group including the plurality of document data as input and outputs the output data, the learning apparatus including: “does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)). “ at least one processor for training, wherein the processor for training is configured to: use each document data for training included in a document data group for training as input of the machine learning model to acquire output data which is output for each document data for training; calculate, for a part of the document data for training from the document data group for training” does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)). “ a loss function representing a degree of difference between correct answer data and the output data for each document data for training based on the output data obtained for each document data for training and the correct answer data; and update the machine learning model based on the loss function” does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)). Step 2B: “ An information processing apparatus comprising: at least one processor, wherein the processor is configured to: “does not amount significantly more than the judicial exception in the claim. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)). “ for a machine learning model that uses a document data group including a plurality of document data as input and outputs output data” does not amount significantly more than the judicial exception in the claim. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)). “ wherein the machine learning model is a machine learning model trained by a learning apparatus of the machine learning model that uses the document data group including the plurality of document data as input and outputs the output data, the learning apparatus including: “does not amount significantly more than the judicial exception in the claim. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)). “ at least one processor for training, wherein the processor for training is configured to: use each document data for training included in a document data group for training as input of the machine learning model to acquire output data which is output for each document data for training; calculate, for a part of the document data for training from the document data group for training” does not amount significantly more than the judicial exception in the claim. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)). “ a loss function representing a degree of difference between correct answer data and the output data for each document data for training based on the output data obtained for each document data for training and the correct answer data; and update the machine learning model based on the loss function” does not amount significantly more than the judicial exception in the claim. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)). In regard to claim 18: Step 2A Prong 1: “ deriving an evaluation value in the machine learning model for each document data included in the document data group” is a mental step of data manipulation. Additional Elements Step 2A Prong 2: “ An information processing method executed by a processor of an information processing apparatus including at least one processor, the information processing method comprising: for a machine learning model that uses a document data group including a plurality of document data as input and outputs output data” does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)). Step 2B: “ An information processing method executed by a processor of an information processing apparatus including at least one processor, the information processing method comprising: for a machine learning model that uses a document data group including a plurality of document data as input and outputs output data” does not amount significantly more than the judicial exception in the claim. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)). In regard to claim 19: Additional Elements Step 2A Prong 2: “ A non-transitory computer-readable medium storing an information processing program that is executable by the processor included in the information processing device to perform the information processing method according to claim 18” does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)). Step 2B: “ A non-transitory computer-readable medium storing an information processing program that is executable by the processor included in the information processing device to perform the information processing method according to claim 18” does not amount significantly more than the judicial exception in the claim. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)). Claim Rejections - 35 USC § 102 07-07-aia AIA 07-07 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 – 07-08-aia AIA (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. Claims 1, 5 and 18 are rejected under 35 U.S.C. 102(a)(1) as being anticipated over Omar Odibat et al. (hereinafter Odibat ) US 2021/0248457 A1. In regard to claim 1 : Odibat discloses: - An information processing apparatus comprising: at least one processor, wherein the processor is configured to: for a machine learning model In [0108]: computer implemented method, system or apparatus, and computer program product are provided in the illustrative embodiments for managing participation in online communities and other related features, functions, or operations. Where an embodiment or a portion thereof is described with respect to a type of device, the computer implemented method, system or apparatus, the computer program product, or a portion thereof, are adapted or configured for use with a suitable and comparable manifestation of that type of device. In [0110]: The present invention may be a system, a method, and/or a computer program product at any possible technical detail level of integration. The computer program product may include a computer readable storage medium (or media) having computer readable program instructions thereon for causing a processor to carry out aspects of the present invention. In [0026]: An illustrated embodiment includes a machine learning system that learns from the assets that a company has already tagged and then accurately predicts tagging information for the remaining untagged assets. The illustrative embodiments recognize that combining one or more feature extraction models with a classification model such that the classification model uses output from the feature extraction model provides for improved accuracy over using the classification model on its own. - that uses a document data group including a plurality of document data as input and outputs output data In [0007]: The illustrative embodiments provide for feature generation for asset classification. An embodiment includes generating an input document that includes a plurality of text fields of attribute data associated with a group of cloud computing assets having a specified label for a specified class. (BRI: the group of cloud computing assets data represents the document data group) In [0034] : In an illustrated embodiment, an application for classifying digital assets receives input data that includes attribute data for cloud computing assets in the form of text and non-text data. In [0029]: the illustrated embodiments allow a user to influence the tradeoff between the accuracy and explainability of the final output. For example, in an embodiment, a user provides an input indicative of a preference for prioritizing accuracy or explainability of the final classifier output. - derive an evaluation value in the machine learning model for each document data included in the document data group In [0007]: the embodiment also includes extracting a set of candidate features from the attribute data using a feature extraction module that evaluates the attribute data using first and second machine learning models, wherein the first machine learning model scores terms in the input document and outputs a first portion of the set of candidate features rank ed according to the term scores, and wherein the second machine learning model includes a deep learning model and outputs a second portion of the set of candidate features ranked according to term probabilities (BRI: the ranking of assets based on term scores is indeed associated to the evaluating their value in document data that involves performance indicators) In [ 0036 ]: In an illustrated embodiment, a modified TF-IDF module creates an input document by merging text fields of assets tagged with a same class label (label-1) to form a single document D(label-1). In some such embodiments, the modified TF-IDF modules calculates scores for all terms in the document. In some such embodiments, the top N most important terms in the context of a class label are chosen based on the highest scores and are placed in a term list term_list(label-1 (BRI: the TF-IDF is a performance indicator ranking based on the importance of specific terms within the documents and within the context of the TF-IDF, the score is the “evaluation value”). In regard to claim 5 : Odibat discloses: wherein the processor is configured to: - normalize each document data included in the document data group; In [0008]: The embodiment also includes calculating feature-selection values for respective features of the set of candidate features , wherein the calculating includes normal izing the candidate feature values and penalizing candidate feature values of candidate features from the second machine learning model based on the user input. In [0009]: Another embodiment includes generating an input document that includes a plurality of text fields of attribute data associated with a group of cloud computing assets having a specified label for a specified class. The embodiment also includes extracting a set of candidate features from the attribute data using a feature extraction module that evaluates the attribute data using first and second machine learning models. - and derive the evaluation value for each normalized document data. In [0007]: The embodiment also includes calculating feature-selection values for respective features of the set of candidate features, wherein the calculating includes normal izing the term scores and ranking the set of candidate features based on a feature-selection algorithm In [0007]: the embodiment also includes extracting a set of candidate features from the attribute data using a feature extraction module that evaluates the attribute data using first and second machine learning models, wherein the first machine learning model scores terms in the input document and outputs a first portion of the set of candidate features rank ed according to the term scores, and wherein the second machine learning model includes a deep learning model and outputs a second portion of the set of candidate features ranked according to term probabilities (BRI: the ranking of assets based on term scores is indeed associated to the evaluating their value in document data that involves performance indicators) In [ 0036 ]: In an illustrated embodiment, a modified TF-IDF module creates an input document by merging text fields of assets tagged with a same class label (label-1) to form a single document D(label-1). In some such embodiments, the modified TF-IDF modules calculates scores for all terms in the document. In some such embodiments, the top N most important terms in the context of a class label are chosen based on the highest scores and are placed in a term list term_list(label-1 (BRI: the TF-IDF is a performance indicator ranking based on the importance of specific terms within the documents and within the context of the TF-IDF, the score is the “evaluation value”). In regard to claim 18 : Odibat discloses: - An information processing method executed by a processor of an information processing apparatus including at least one processor, the information processing method comprising: In [0108]: computer implemented method, system or apparatus, and computer program product are provided in the illustrative embodiments for managing participation in online communities and other related features, functions, or operations. Where an embodiment or a portion thereof is described with respect to a type of device, the computer implemented method, system or apparatus, the computer program product, or a portion thereof, are adapted or configured for use with a suitable and comparable manifestation of that type of device. In [0110]: The present invention may be a system, a method, and/or a computer program product at any possible technical detail level of integration. The computer program product may include a computer readable storage medium (or media) having computer readable program instructions thereon for causing a processor to carry out aspects of the present invention. In [0026]: An illustrated embodiment includes a machine learning system that learns from the assets that a company has already tagged and then accurately predicts tagging information for the remaining untagged assets. The illustrative embodiments recognize that combining one or more feature extraction models with a classification model such that the classification model uses output from the feature extraction model provides for improved accuracy over using the classification model on its own. - for a machine learning model that uses a document data group including a plurality of document data as input and outputs output data, In [0007]: The illustrative embodiments provide for feature generation for asset classification. An embodiment includes generating an input document that includes a plurality of text fields of attribute data associated with a group of cloud computing assets having a specified label for a specified class. (BRI: the group of cloud computing assets data represents the document data group) In [0034] : In an illustrated embodiment, an application for classifying digital assets receives input data that includes attribute data for cloud computing assets in the form of text and non-text data. In [0029]: the illustrated embodiments allow a user to influence the tradeoff between the accuracy and explainability of the final output. For example, in an embodiment, a user provides an input indicative of a preference for prioritizing accuracy or explainability of the final classifier output. - deriving an evaluation value in the machine learning model for each document data included in the document data group. In [0007]: the embodiment also includes extracting a set of candidate features from the attribute data using a feature extraction module that evaluates the attribute data using first and second machine learning models, wherein the first machine learning model scores terms in the input document and outputs a first portion of the set of candidate features rank ed according to the term scores, and wherein the second machine learning model includes a deep learning model and outputs a second portion of the set of candidate features ranked according to term probabilities (BRI: the ranking of assets based on term scores is indeed associated to the evaluating their value in document data that involves performance indicators) In [ 0036 ]: In an illustrated embodiment, a modified TF-IDF module creates an input document by merging text fields of assets tagged with a same class label (label-1) to form a single document D(label-1). In some such embodiments, the modified TF-IDF modules calculates scores for all terms in the document. In some such embodiments, the top N most important terms in the context of a class label are chosen based on the highest scores and are placed in a term list term_list(label-1 (BRI: the TF-IDF is a performance indicator ranking based on the importance of specific terms within the documents and within the context of the TF-IDF, the score is the “evaluation value”). Claims 10-11, 13-16, and 20-21 are rejected under 35 102(a)(1) as being anticipated over Alex Londeree et al. (hereinafter Londer ) US 2023/0134791 A1. In regard to claim 10 : Londer discloses: - A learning apparatus of a machine learning model that uses a plurality of document data as input and outputs output data, the learning apparatus comprising: In [0060]: The data processing apparatus described herein with respect to FIGS. 1 - 8 may be used to facilitate operation of a QA system that provides document level inference according to some embodiments of the inventive concept described herein. These apparatus may be embodied as one or more enterprise, application, personal, pervasive and/or embedded computer systems and/or apparatus that are operable to receive, transmit, process and store data In [0045] : The retriever engine 205 is further configured to encode both the query and the sub-documents . In some embodiments, the retriever engine 205 may use a vector space model to encode the query and/or the sub-documents (BRI: query and sub-documents are inputs) In [0044]: the retriever engine 205 may be configured to divide the document in the knowledge base 210 into a plurality of sub-documents, which may be indexed based on a suitable categorization indicium. In [0045]: classification involves prediction of an output from a set of finite categorical values. - at least one processor, wherein the processor is configured to: use, for a plurality of document data for training, each document data for training as input of the machine learning model to acquire output data which is output for each document data for training; In [0044]: During training, the retriever engine 205 may be configured to discard a current knowledge base or corpus and select a new knowledge base or corpus 215 that may include one or more of the plurality of sub-documents . In [0045]: During training, the reader engine 230 may be configured to generate an inference about the answer to the query based on each of the matching sub-documents that are concatenated with the query. - calculate, for a part of the document data for training from the plurality of document data for training, a loss function representing a degree of difference between correct answer data and the output data for each document data for training based on the output data obtained for each document data for training and the correct answer data; In [0045]: the retriever engine 205 may be configured to update the query encoding model and the document encoding model using a match loss function that generates a match loss function result based on the query and the matching sub-documents. A loss function may be described generally as a function that computes the distance between the current output of an operation and the expected output . In accordance with various embodiments of the inventive concept, the reader loss function and/or the match loss function may comprise a regression loss function or a classification loss function. (BRI: the distance is the degree of similarity) - and update the machine learning model based on the loss function. In [0034]: the query encoding model and the document encoding model may be updated using a match loss function result between the query and the matching sub-documents . In [0045]: reader loss function may be applied to identify the matching sub-document having the lowest reader loss function result. This identified matching sub-document may then be associated with a truth label for the query and the AI reader engine 230 may be update d based on the reader loss function results associated with the matching sub-documents. In regard to claim 11 : Londer discloses: - wherein the processor is configured to extract the part of document data for training based on a degree of similarity between the output data and the correct answer data In [0039] : According to some embodiments of the inventive concept, a QA system may be provided to assist entities, such as providers, payors, researchers, and others to perform queries using patient health care records as a knowledge base or corpus. The QA system may include a health care facility interface server 130 , which includes an EMR interface system module 135 to facilitate the transfer of information between the EMR system 120, which the providers use to manag e patient charts and records and issue orders In [0039]: The QA system server 140 along with the QA engine module 145 may be configured to embody an AI driven retriever engine and reader engine for generating inferences in response to one or more queries using a knowledge base or corpus that may be based on the patient records and/or other health care or medical information. In [0045]: the retriever engine 205 may be configured to update the query encoding model and the document encoding model using a match loss function that generates a match loss function result based on the query and the matching sub-documents. A loss function may be described generally as a function that computes the distance between the current output of an operation and the expected output (BRI : the QA system is a part of the document data to manages the patient charts to ensure completeness and accuracy of data) In regard to claim 13 : Londer discloses: - wherein the processor is configured to calculate, for the part of document data for training, the loss function by performing weighting based on the output data obtained for each document data for training and the correct answer data. In [0045]: A vector space model is an algebraic model for representing text documents as vectors of identifiers. Each vector dimension corresponds to a separate term. If a term occurs in the document, its value in the vector is non-zero. There are multiple ways to compute the term values including using a Term Frequency-Inverse Document Frequency (TF-IDF) weighting scheme . TF-IDF is a numerical statistic that is designed to reflect how important a word is to a document in a collection or corpus In regard to claim 14: Londer discloses: - wherein the processor is configured to set weighting to be larger as a degree of similarity between the output data and the correct answer data is higher. In [0035]: A query is received and the retriever engine encodes both the query and the sub-documents. The retriever engine determines a sub-document that is a best match for containing the answer to the query using, for example, a match loss function . The query and the matching sub-document are concatenated and communicated to the reader engine, which generates an inference about the answer to the query based on the concatenated matching sub-document and query. In [ 0045 ]: The retriever engine 205 is further configured to encode both the query and the sub-documents . In [0045]: A loss function may be described generally as a function that computes the distance between the current output of an operation and the expected output In [0045] : There are multiple ways to compute the term values including using a Term Frequency-Inverse Document Frequency (TF-IDF) weighting scheme. TF-IDF is a numerical statistic that is designed to reflect how important a word is to a document in a collection or corpus In [0045]: During training, the reader engine 230 may be configured to generate an inference about the answer to the query based on each of the matching sub-documents that are concatenated with the query. A reader loss function may be applied to identify the matching sub-document having the lowest reader loss function result . This identified matching sub-document may then be associated with a truth label for the query and the AI reader engine 230 may be updated based on the reader loss function results associated with the matching sub-documents. Similarly, the retriever engine 205 may be configured to update the query encoding model and the document encoding model using a match loss function that generates a match loss function result based on the query and the matching sub-documents. (BRI: a lowest loss function reflects on highest degree of similarity) In [0045]: A loss function may be described generally as a function that computes the distance between the current output of an operation and the expected output. In accordance with various embodiments of the inventive concept, the reader loss function and/ or the match loss function may comprise a regression loss function or a classification loss function. Broadly described, classification involves prediction of an output from a set of finite categorical values . Regression involves prediction of a continuous value from other information. In regard to claim 15 : Londer discloses: wherein the processor is configured to: - repeatedly update the machine learning model based on the loss function obtained from the part of document data for training; In [ 0034 ] : Some embodiments of the inventive concept differ from conventional QA systems that are trained using a single large knowledge base or corpus dataset to make inferences in that each training iteration and each inference replaces the previous knowledge base or corpus with a unique document. In [0034]: the QA system according to embodiments of the inventive concept may use a retriever engine-reader engine architecture in which the retriever engine discards a current knowledge base or corpus, selects a new knowledge base or corpus, which is divided into multiple sub-documents. During training mode, both a query and the sub-documents may be encoded and one or more matching sub-documents may be determined based on the encoded query and sub-documents using the retriever engine . The matching sub-documents are each concatenated with the query and communicated to the reader engine, which generates an inference about the answer to the query based on each of these sub-documents that are concatenated with the query. A reader loss function is applied to identify the matching sub-document having the lowest reader loss function result. In [0034]: the query encoding model and the document encoding model may be update d using a match loss function result between the query and the matching sub-documents. - and change the number of the part of document data for training to be extracted, according to the number of updates of the machine learning model. in [0044]: select a new knowledge base or corpus 215 that may include one or more of the plurality of sub-documents . In regard to claim 16 : Londer discloses: - wherein: each document data for training is given with a label representing a type of an associated prediction result of the machine learning model, 0045, 0048 In [ 0045]: In accordance with various embodiments of the inventive concept, the reader loss function and/or the match loss function may comprise a regression loss function or a classification loss function. Broadly described, classification involves prediction of an output from a set of finite categorical values. Regression involves prediction of a continuous value from other information. In [ 0045 ]; The retriever engine 205 is further configured to encode both the query and the sub-documents. In some embodiments, the retriever engine 205 may use a vector space model to encode the query and/or the sub-documents In [0003]: AI reader engine: generating an inference about the answer to the query based on a concatenation of each of the at least one matching sub-document with the query, each of the at least one matching sub-document having an associated reader loss function result for the inference; identifying one of the at least one matching sub-document having a lowest reader loss function result; and associating the identified one of the at least one matching sub-document with a truth label for the query. - and the processor is configured to extract the document data for training for each type of the label. In regard to claim 20 : Londer discloses: - A learning method comprising: via a processor, using, for a plurality of document data for training, each document data for training as input of a machine learning model to acquire output data which is output for each document data for training; In [0044]: During training, the retriever engine 205 may be configured to discard a current knowledge base or corpus and select a new knowledge base or corpus 215 that may include one or more of the plurality of sub-documents . In [0045]: During training, the reader engine 230 may be configured to generate an inference about the answer to the query based on each of the matching sub-documents that are concatenated with the query. - calculating, for a part of the document data for training from the plurality of document data for training, a loss function representing a degree of difference between correct answer data and the output data for each document data for training based on the output data obtained for each document data for training and the correct answer data; In [0045]: the retriever engine 205 may be configured to update the query encoding model and the document encoding model using a match loss function that generates a match loss function result based on the query and the matching sub-documents. A loss function may be described generally as a function that computes the distance between the current output of an operation and the expected output . In accordance with various embodiments of the inventive concept, the reader loss function and/or the match loss function may comprise a regression loss function or a classification loss function. (BRI: the distance is the degree of similarity) - and updating the machine learning model based on the loss function. In [0034]: the query encoding model and the document encoding model may be updated using a match loss function result between the query and the matching sub-documents . In [0045]: reader loss function may be applied to identify the matching sub-document having the lowest reader loss function result. This identified matching sub-document may then be associated with a truth label for the query and the AI reader engine 230 may be update d based on the reader loss function results associated with the matching sub-documents. In regard to claim 21 : Londer discloses: - A non-transitory computer-readable medium storing a learning program that is executable by the processor included in an information processing device to perform the learning method according to claim 20. In [0022]: Claim Rejections - 35 USC § 103 07-20-aia AIA 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 2019to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made. 07-21-aia AIA Claim s 2, 7-8 are rejected under 35 U.S.C. 103 as being unpatentable over Omar Odibat et al. (hereinafter Odibat ) US 2021/0248457 A1. further In view of Weiguo Li et al. (hereinafter Li ) US 2019/0370304 A1. In regard to claim 2 : Odibat does not explicitly disclose: - wherein the processor is configured to perform at least one of specification of the document data, which is a display target, from the document data group or specification of a display order of a document according to the document data based on the derived evaluation value. However, Li discloses: - wherein the processor is configured to perform at least one of specification of the document data, which is a display target, from the document data group or specification of a display order of a document according to the document data based on the derived evaluation value. In [0010]: In a feasible embodiment, before the ranking, by the server , the N documents by using a ranking model to obtain a document ranking list , the method includes obtaining, by the server, the historical search word, and obtaining M documents corresponding to the historical search word; ranking, by the server, the M documents to obtain a target document ranking list; In [0040]: the server ranks the N documents by using a ranking model to obtain a document ranking list specifica lly includes scoring , by the server, each of the N documents based on the ranking model, and obtaining a scoring result; and ranking, by the server, the N documents based on the scoring result to obtain the document ranking list. (BRI: a document ranking list represents a document order) In [0043]: S104: The server display s the document ranking list to the user . It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Odibat and LI . Odibat teaches machine learning based document processing providing an evaluation value for the plurality of the documents data. Li teaches word extraction and document ranking. One of ordinary skill would have motivation to combine Odibat and Li that can improve the user satisfaction of ranking result of the document ( Li [0011]) . In regard to claim 7 : Odibat discloses: wherein: - and the processor is configured to use each combination data in which the first document data and the second document data are combined as input of the machine learning model to derive a second evaluation value from output data which is output for each combination data. In [0008]: Another embodiment includes generating an input document that includes a plurality of text fields of attribute data associated with a group of cloud computing assets having a specified label for a specified class. The embodiment also includes extracting a set of candidate features from the attribute data using a feature extraction module that evaluates the attribute data using first and second machine learning models that output respective first and second portions of the set of candidate features having respective candidate feature values Odibat does not explicitly disclose: - the document data having a greatest first evaluation value, which is derived for each document data, is used as first document data, and each of the plurality of document data other than the first document data included in the document data group is used as second document data, However, Li discloses: - the document data having a greatest first evaluation value, which is derived for each document data, is used as first document data, and each of the plurality of document data other than the first document data included in the document data group is used as second document data, In [0013]: In a feasible embodiment, the obtaining, by the server, a ranking effect evaluation value of the target document ranking list includes evaluating, by the server, a ranking effect of the target document ranking list based on a user behavior, and obtaining the ranking effect evaluation value. In [0014]: In a feasible embodiment, the obtaining, by the server, a ranking effect evaluation value of the target document ranking list further includes using, by the server as the ranking effect evaluation value, a value provided by the user after the user evaluates a ranking effect of the target document ranking list. In [0012]: In a feasible embodiment, the ranking, by the server, the M documents based on a ranking model to obtain a target document ranking list includes scoring , by the server, a correlation between each of the M documents and the historical search word based on the ranking model to obtain a scoring result; and ranking, by the server, the M documents in ascending order or descending order of the scoring results to obtain the target document ranking list. In [0053]: The server ranks the M documents in ascending order or descending order of the scoring results to obtain the document ranking list. In [0055] : Scoring results obtained by the server by scoring occurrence frequency of the historical search word in all of the M documents based on the ranking model f (q, d, θ) may be represented by a set (d.sub.1′, d.sub.2′ , . . . , d.sub.M′), and the target document ranking list obtained by the server by ranking the M documents in ascending order or descending order of the scoring results may be represented by a set (y.sub.1, y.sub.2, . . . , y.sub.M). In [0056]: Further, the foregoing process can be represented by σ=(y.sub.1, y.sub.2, . . . , y.sub.M)=sort (d.sub.1, d.sub.2, . . . , d.sub.M). The sort function is a descending-order ranking model or an ascending-order ranking model. (BRI: Example: scores {72, 75, 82 and 88} ranked using descending order {88, 82, 75 , 72}. The document with score 88 represents highest evaluation value recognized as “first document data”(88) and other document next to it as “second document data”) It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Odibat and Li. Odibat teaches machine learning based document processing providing an evaluation value for the plurality of the documents data. Li teaches the largest evaluation value. One of ordinary skill would have motivation to combine Odibat and Li that can improve the user satisfaction of ranking result of the document ( Li [0011]) . In regard to claim 8 : Odibat does not explicitly disclose: wherein the processor is configured to: - give a first display priority to the first document data; - and give a second display priority, which is lower than the first display priority, to the second document data based on the second evaluation value. However, Li discloses: wherein the processor is configured to: - give a first display priority to the first document data; In [0008]: receiving, by a server , a search word entered by a user; obtaining, by the server, N documents matching the search word, wherein N is a natural number; ranking, by the server, the N documents by using a ranking model to obtain a document ranking list, in [0012]: In a feasible embodiment, the ranking, by the server, the M documents based on a ranking model to obtain a target document ranking list includes scoring, by the server, a correlation between each of the M documents and the historical search word based on the ranking model to obtain a scoring result; and ranking, by the server, the M documents in ascending order or descending order of the scoring results to obtain the target document ranking list. In [0008]: and display ing, by the server, the document ranking list to the user. (BRI: the server displays the ranking as a first priority when the ranking is sorted in descending order) - and give a second display priority, which is lower than the first display priority, to the second document data based on the second evaluation value. In [0008]: receiving, by a server , a search word entered by a user; obtaining, by the server, N documents matching the search word, wherein N is a natural number; ranking, by the server, the N documents by using a ranking model to obtain a document ranking list, in [0012]: In a feasible embodiment, the ranking, by the server, the M documents based on a ranking model to obtain a target document ranking list includes scoring, by the server, a correlation between each of the M documents and the historical search word based on the ranking model to obtain a scoring result; and ranking, by the server, the M documents in ascending order or descending order of the scoring results to obtain the target document ranking list. In [0008]: and display ing, by the server, the document ranking list to the user. (BRI: the server displays the ranking as a second priority when the ranking is sorted in ascending order in which the lowest ranked is at the top) It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Odibat and Li. Odibat teaches machine learning based document processing providing an evaluation value for the plurality of the documents. Li teaches the largest evaluation value. One of ordinary skill would have motivation to combine Odibat and Li that can improve the user satisfaction of ranking result of the document ( Li [0011]) . Claims 3-4, 6 and 9 are rejected under 35 103 as being unpatentable over Omar Odibat et al. (hereinafter Odibat ) US 2021/0248457 A1. further in view of Aurélien COQUARD et al . (hereinafter COQUARD) ) US 2020/0327172 A1. In regard to claim 3 : Odibat does not explicitly disclose: - wherein the processor is configured to: use each document data as input of the machine learning model to acquire document unit output data which is output for each document data; However, COQUARD discloses: - wherein the processor is configured to: use each document data as input of the machine learning model to acquire document unit output data which is output for each document data; In [ 0012 ]: there is provided a system for processing a plurality of contract documents having different formats and clauses . The system comprises at least one processor programmed or configured to: search contract documents to form one or more groups of contract documents by selecting a first contract document for the or each group and searching for other contract documents having a relevance score within a relevance threshold ; determine a most recently revised contract document within the or each group and determining similarity score for each contract document in said group against the most recently revised contract document for the group in [0095]: The modeling engine 104 processes the clauses separately to generate a representation vector for each clause. A representation vector for a clause may be any number of dimensions and may be based on a sequence of word embeddings a nd/or a sequence of sentence embeddings. In examples, the modeling engine 104 employs an embedding model that is generated from processing textual documents including, but not limited to, a plurality of contract documents, news articles, webpages, and/or any other like text. The model may be a neural network that is trained from such training texts . - and derive the evaluation value for each document data based on the document unit output data. In [0114]: In some embodiments, the comparison of representation vectors may be evaluated based on an unsupervised metric that does not require any labels or ground truth data. For example, the metric may be a percentage of character matches based on a semantic differential. The metric may increase each time a closer (i .e., shorter distance ) clause is found. Such a metric may be used to evalua te the quality of the embedding model and/or algorithms for parsing contract documents, classifying clauses, and/or the like. It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Odibat and COQUARD. Odibat teaches machine learning based document processing providing an evaluation value for the plurality of the documents. COQUARD teaches document unit data and its evaluation. One of ordinary skill would have motivation to combine Odibat and COQUARD t hat provide a shorter distance as a performance metric ( COQUARD [0114]). In regard to claim 4 : Odibat discloses: - wherein the evaluation value has a correlation with the document unit output data. In [0034]: In an illustrated embodiment, an application for classifying digital assets receives input data that includes attribute data for c loud computing assets in the form of text and non-text dat a. A non-limiting example of input data includes data extracted from a bill or invoice or other data source for cloud assets computing that includes cost and usage information from one or more providers of cloud computing assets, as well as other data representative of various attribute information that can be mapped to a list of generic attribute columns, such as the following non-limiting example, which is included as an example only for clarity purposes: PNG media_image1.png 272 395 media_image1.png Greyscale (BRI: See line “correlation _asset id” under generic asset data in the Table above” that represents the correlation of document unit output data with the evaluation value provided as TF-IDF ) In [0023]: The number of cloud computing assets in many organizations continues to grow, and for some organizations the number of assets is quickly growing to, or has already surpassed, a point where it is impractical for users to manually manage so many assets. In [0024]: Management of digital assets is further complicated by the fact that organizations within many companies have the authority to independently provision and configure digital assets from multiple different providers . To further complicate things, each provider typically has a different and unique process for tagging and tracking digital assets; some provisioned assets are repurposed and reclassified as organizations shift their strategic focus; and new digital assets arise quickly because many cloud providers include automatic software installation and configuration as a standard part of system deployment. In [0025]: the illustrative embodiments also recognize that a cognitive asset management system that provides accurate recognition and classification results would be useful for automating many tasks associated with the management of digital assets. In [0062]: daata processing environment 100 may also take the form of a cloud, and employ a cloud computing model of service delivery for enabling convenient, on-demand network access to a shared pool of configurable computing resources (e.g. networks, network bandwidth, servers, processing, memory, storage, applications, virtual machines, and services) that can be rapidly provisioned and released with minimal management effort or interaction with a provider of the service. (BRI: Asset management system data is a “document unit data”) In [0035]: in an embodiment, the feature extraction module includes a deep learning model and a modified TF-IDF model for receiving text input data and standard feature engineering and aggregations for non-text input data . In an embodiment, the output from the feature extraction module includes a plurality of candidate features that are provided to a customizable feature selection module. In an embodiment, the customizable feature selection module ranks the candidate features in order of importance. In [0036]: In an illustrated embodiment, a modified TF-IDF module creates an input document by merging text fields of assets tagged with a same class label (label-1) to form a single document D(label-1). In some such embodiments, the modified TF-IDF modules calculates scores for all terms in the document. In regard to claim 6 : Od i bat discloses: - the evaluation value in the machine learning model as a word unit evaluation value for each word data; In [0026]: An illustrated embodiment includes a machine learning system that learns from the assets that a company has already tagged and then accurately predicts tagging information for the remaining untagged assets. In [0032]: The illustrative embodiments further recognize that the use of a TF-IDF model makes it easier to understand how the occurrence of a specific word in a text field contributes to the asset being associated with a class label. However, accuracy is also a key goal, so the illustrative embodiments recognize that a modified TF-IDF is more accurate if it is more closely associated with a desired context by modifying the TF-IDF to re-define an input document to the TF-IDF. The illustrative embodiments also recognize that a modified TF-IDF is more accurate if the IDF component is revised to include term frequency information in place of a binary indication of a number of documents containing the term. (BRI: TF-IDF represents a “word unit evaluation value”) - and derive the evaluation value according to a statistical value of the word unit evaluation value of the word data included in the document data for each document data. In [0007]: the embodiment also includes extracting a set of candidate features from the attribute data using a feature extraction module that evaluates the attribute data using first and second machine learning models, wherein the first machine learning model scores terms in the input document and outputs a first portion of the set of candidate features rank ed according to the term scores, and wherein the second machine learning model includes a deep learning model and outputs a second portion of the set of candidate features ranked according to term probabilities (BRI: the ranking of assets based on term scores is indeed associated to the evaluating their value in document data that involves performance indicators) In [ 0036 ]: In an illustrated embodiment, a modified TF-IDF module creates an input document by merging text fields of assets tagged with a same class label (label-1) to form a single document D(label-1). In some such embodiments, the modified TF-IDF modules calculates scores for all terms in the document. In some such embodiments, the top N most important terms in the context of a class label are chosen based on the highest scores and are placed in a term list term_list(label-1 (BRI: the TF-IDF is statistical measure and is a performance indicator ranking based on the importance of specific terms within the documents and within the context of the TF-IDF, the score is the “evaluation value”). (BRI: TF-IDF is an evaluation value) Odibat does not explicitly disclose: - extract a plurality of word data from each document data included in the document data group; However, COQUARD discloses: - extract a plurality of word data from each document data included in the document data group; In [0011]: the method comprises parsing a first contract document to identify a plurality of clauses in the first contract document , each clause of the plurality of clauses comprising a sequence of word s ; generating a plurality of representation vectors based on the first contract document and at least one embedding model, wherein each representation vector of the plurality of representation vectors is generated based on a separate clause of at least a subset of clauses of the plurality of clauses It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Odibat and COQUARD. Odibat teaches machine learning based document processing providing an evaluation value for the plurality of the documents and statistical method, COQUARD teaches word extraction. One of ordinary skill would have motivation to combine Odibat and COQUARD that provided shorter distance as a metric [COQUARD [0114]) . In regard to claim 9 : Odibat does not explicitly disclose: - extract a plurality of word data from each document data included in the document data group; - derive the evaluation value in the machine learning model as a word unit evaluation value for each word data - derive a first statistical value of the word unit evaluation value of the word data included in the document data for each document data to give a first evaluation value to first evaluation value document data which is the document data having a greatest first statistical value; - derive, for a plurality of combination data in which the first evaluation value document data, and each of the plurality of document data other than the first evaluation value document data included in the document data group are combined, a second statistical value of the word unit evaluation value of the word data included in the combination data for each combination data to give a second evaluation value, which is lower than the first evaluation value, to second evaluation value document data which is the document data having a greatest second statistical value; - and set, in derivation of the second statistical value, the word unit evaluation value of the word data included in the first evaluation value document data among the word data included in the document data combined with the first evaluation value document data to be relatively lower than the word unit evaluation value of the word data which is not included in the first evaluation value document data. However, COQUARD discloses: - extract a plurality of word data from each document data included in the document data group; In [0011]: In an embodiment, the method comprises parsing a first contract document to identify a plurality of clauses in the first contract document, each clause of the plurality of clauses comprising a sequence of words; generating a plurality of representation vectors based on the first contract document and at least one embedding model, wherein each representation vector of the plurality of representation vectors is generated based on a separate clause of at least a subset of clauses of the plurality of clauses - derive the evaluation value in the machine learning model as a word unit evaluation value for each word data In [0114] : the comparison of representation vectors may be evalua ted based on an unsupervised metric that does not require any labels or ground truth data . For example, the metric may be a percentage of character matches based on a semantic differential . The metric may increase each time a closer (i.e., shorter distance) clause is found. Such a metric may be used to evalua te the quality of the embedding model and/or algorithms for parsing contract documents, classifying clauses, and/or the like. In [0134]: In some embodiments, natural language processing techniques may be utilized to process questions inputted by users about a particular clause or contract document . For example, a linear regression model may be developed based on the word embeddings and/or sentence embeddings to enable automatic determinations of answers to inputted questions. (BRI: the metric is the evaluation value) - derive a first statistical value of the word unit evaluation value of the word data included in the document data for each document data to give a first evaluation value to first evaluation value document data which is the document data having a greatest first statistical value In [0013] : In an embodiment, the relevance score is a word frequency statistic measurement and the similarity score is a word dissimilarity measure . For example, the word frequency statistic measurement may be a term frequency-inverse document frequency value and the word dissimilarity measure may be an edit distance. In [0148]: At 1104, the method selects a next ungrouped document. This may be derived from the already received search result and may be based on any suitable metric such as a next database index number. A relevance score between the two selected documents is also determined. This may be implemented by the relevance score engine 1026A, for example by determining a tf-idf value between the two selected documents. In [0149: At 1106, the method determines whether this relevance score is within a threshold, for example the tf-idf value is greater than 50%. If the relevance score is not within the threshold (N), the method moves to 1110, otherwise (Y) the method moves to 1108. In [0153]: At 1112, all documents having a sufficiently high relevance score based on the first selected document have been identified and added to the group based on the first selected document . The method then identifies and selects the most recently revised contract document in the recently created group. This may be implemented by the search engine searching for the document within the group having the most recent edit date. (BRI: Highest relevance score relative to the selected document provide greatest statistical value In [0158]: The method may be arranged such that the similarity score is determined with respect to the first selected document and all other documents in the group and the results ordered or sorted in descending order of similarity so that those having a similarity score above the threshold are readily identified without having to proceed through all steps of the method for all documents. - derive, for a plurality of combination data in which the first evaluation value document data, and each of the plurality of document data other than the first evaluation value document data included in the document data group are combined In [0117]: In other embodiments, the clause titles may be combined with the clause bodies for generating a representation vector that represents both the title and the clause. (BRI: Within the context of plurality of document data, the examiner interprets the limitation as each document data is assessed separately to determine the document contribution to the overall evaluation. Combining clause titles and bodies can indeed represent combining the evaluation of each document data. Such combination provides better organization of the content but does not change the legal implication of the content In [0035]: In an embodiment, the output data comprises a second contract document , and wherein generating the second contract document comprises determining a counter- proposal to at least one clause of the plurality of clauses based on a contract database comprising a plurality of contract documents. In [0020]: representation vector corresponding to the clause, the output data comprising at least one of the following: a new parameter replacing the parameter, a new clause replacing the clause , an annotation identifying the parameter, an annotation identifying the clause, risk assess ment data based on the parameter , or any combination thereof. In [0130]: the output data may include a contract summary, an annotated contract document, a modified contract document with one or more new parameters, a modified contract document with one or more new clauses, risk assess ment data , a data structure representing a plurality of extracted parameters, a structured contract document based on the first contract document and including merge fields corresponding to a plurality of predetermined fields , metadata to be associated with the first contract document, and/or the like. (BRI: the combination represents evaluating proposals against the assessment of the offer to perform contract successfully. A risk indicate not being successful) - a second statistical value of the word unit evaluation value of the word data included in the combination data for each combination data to give a second evaluation value, which is lower than the first evaluation value, to second evaluation value document data which is the document data having a greatest second statistical value; In [0153]: At 1112, all documents having a sufficiently high relevance score based on the first selected document have been identified and added to the group based on the first selected document. The method then identifies and selects the most recently revised contract document in the recently created group. This may be implemented by the search engine searching for the document within the group having the most recent edit date. In [0158]: The method may be arranged such that the similarity score is determined with respect to the first selected document and all other documents in the group and the results ordered or sorted in descending order of similarity so that those having a similarity score above the threshold are readily identified without having to proceed through all steps of the method for all documents. (A sorted descending order provides lower score for the second evaluation value) - and set, in derivation of the second statistical value, the word unit evaluation value of the word data included in the first evaluation value document data among the word data included in the document data combined with the first evaluation value document data to be relatively lower than the word unit evaluation value of the word data which is not included in the first evaluation value document data. In [0158] : The method may be arranged such that the similarity score is determined with respect to the first selected document and all other documents in the group and the results ordered or sorted in descending order of similarity so that those having a similarity score above the threshold are readily identified without having to proceed through all steps of the method for all documents. (BRI: the similarity score sorted in descending order indeed indicates the second statistical value as lower than the first in which the ranking measure is the resemblance between two documents). It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Odibat and COQUARD. Odibat teaches machine learning based document processing providing an evaluation value for the plurality of the documents and statistical method, COQUARD teaches word extraction. One of ordinary skill would have motivation to combine Odibat and COQUARD that provided shorter distance as a metric [COQUARD [0114]) . 07-21-aia AIA Claim 12 is rejected under 35 U.S.C. 103 as being unpatentable over Alex Londeree et al. (hereinafter Londer ) US 2023/0134791 A1. further in view of Craig Schmuga r et al. (hereinafter Schmugar ) US 2022/0131896 A1. In regard to claim 12 : Londer discloses: - wherein the processor is configured to: calculate, also for another document data for training other than the part of document data for training, In [0012]: In some embodiments of the inventive concept, a method comprises: discarding a current knowledge corpu s; selecting a new knowledge corpus; performing operations as follows using an Artificial Intelligence (AI) retriever engine: dividing the new knowledge corpus into a plurality of sub-documents ; encoding a query for the plurality of sub-documents using a query encoding model; encoding each of the plurality of sub-documents using a document encoding model; and determining one matching sub-document of the plurality of sub-documents that is a best match for containing an answer to the query based on the encoded query and each of the plurality of encoded sub-documents ; performing operations as follows using an AI reader engine: generating an inference about the answer to the query based on a concatenation of the one matching sub-document with the query. (BRI: the new knowledge training does represent other document training) Londer does not explicitly disclose: - the loss function with a weight smaller than a weight of the part of document data for training for each data for training; - and update the machine learning model based also on the loss function of the other document data for training. Londer 0034 However, Schmugar discloses: - the loss function with a weight smaller than a weight of the part of document data for training for each data for training; In [0028]: within an industry consortium, it may be expected to share certain enterprise data across the consortium. However, it may be more common to share the enterprise data in a static format such as portable document format (PDF) or a tagged image file (TIF), rather than in a native electronic format such as a Microsoft Office document, or other native binary data. In [ 0267 ]: A common method for refining values includes “gradient descent” and “back-propagation.” An illustrative gradient descent method includes computing a “cost” function, which measures the error in the network. In [0267]: Each training example will have its own computed cost. Initially, the cost function is very large, because the network doesn't know how to classify objects. As the network is trained and refined, the cost function value is expected to get smaller, as the weights and biases are adjusted toward more useful values. ( BRI: smaller weight means that a loss function has given more more importance to certain data points during training, which may require a weighted loss function . This modifies the standard loss by multiplying each term by a class weight or sample weight . - and update the machine learning model based also on the loss function of the other document data for training. In [0283] : Individual cost functions can be computed for each training input, and the total cost function for the network can be computed as an average of the individual cost functions. In [0284]: In block 1420, the network may then compute a negative gradient of this cost function to seek a local minimum value of the cost function, or in other words, the error. For example, the system may use back-propagation to seek a negative gradient numerically. After computing the negative gradient, the network may adjust parameters (weights and biases) by some amount in the “downward” direction of the negative gradient. It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Odibat and Schmugar. Odibat teaches word extraction and statistical method. Schmugar teaches the loss function with a smaller weight (optimized weight) One of ordinary skill would have motivation to combine Odibat and Schmugar that can adjustment of the slope of the cost function 0271]) . 07-21-aia AIA Claim s 17 and 19 are rejected under 35 U.S.C. 103 as being unpatentable over Omar Odibat et al. (hereinafter Odibat ) US 2021/0248457 A1. Further in view of Alex Londeree et al. (hereinafter Londer ) US 2023/0134791 A1. In regard to claim 17 : Odibat discloses: An information processing apparatus comprising: - at least one processor, wherein the processor is configured to: for a machine learning model that uses a document data group including a plurality of document data as input and outputs output data, In [0108]: computer implemented method, system or apparatus, and computer program product are provided in the illustrative embodiments for managing participation in online communities and other related features, functions, or operations. In [0110]: The present invention may be a system, a method, and/or a computer program product at any possible technical detail level of integration. The computer program product may include a computer readable storage medium (or media) having computer readable program instructions thereon for causing a processor to carry out aspects of the present invention. In [0026]: An illustrated embodiment includes a machine learning system that learns from the assets that a company has already tagged and then accurately predicts tagging information for the remaining untagged assets. The illustrative embodiments recognize that combining one or more feature extraction models with a classification model such that the classification model uses output from the feature extraction model provides for improved accuracy over using the classification model on its own. In [0007]: The illustrative embodiments provide for feature generation for asset classification. An embodiment includes generating an input document that includes a plurality of text fields of attribute data associated with a group of cloud computing assets having a specified label for a specified class. (BRI: the group of cloud computing assets data represents the document data group) In [0034] : In an illustrated embodiment, an application for classifying digital assets receives input data that includes attribute data for cloud computing assets in the form of text and non-text data. In [0029]: the illustrated embodiments allow a user to influence the tradeoff between the accuracy and explainability of the final output. For example, in an embodiment, a user provides an input indicative of a preference for prioritizing accuracy or explainability of the final classifier output. - derive an evaluation value in the machine learning model for each document data included in the document data group, In [0007]: the embodiment also includes extracting a set of candidate features from the attribute data using a feature extraction module that evaluates the attribute data using first and second machine learning models, wherein the first machine learning model scores terms in the input document and outputs a first portion of the set of candidate features rank ed according to the term scores, and wherein the second machine learning model includes a deep learning model and outputs a second portion of the set of candidate features ranked according to term probabilities (BRI: the ranking of assets based on term scores is indeed associated to the evaluating their value in document data that involves performance indicators) In [ 0036 ]: In an illustrated embodiment, a modified TF-IDF module creates an input document by merging text fields of assets tagged with a same class label (label-1) to form a single document D(label-1). In some such embodiments, the modified TF-IDF modules calculates scores for all terms in the document. In some such embodiments, the top N most important terms in the context of a class label are chosen based on the highest scores and are placed in a term list term_list(label-1 (BRI: the TF-IDF is a performance indicator ranking based on the importance of specific terms within the documents and within the context of the TF-IDF, the score is the “evaluation value”). - wherein the machine learning model is a machine learning model trained by a learning apparatus of the machine learning model that uses the document data group including the plurality of document data as input and outputs the output data, In [0026]: An illustrated embodiment includes a machine learning system that learns from the assets that a company has already tagged and then accurately predicts tagging information for the remaining untagged assets. The illustrative embodiments recognize that combining one or more feature extraction models with a classification model such that the classification model uses output from the feature extraction model provides for improved accuracy over using the classification model on its own. In [0007]: The illustrative embodiments provide for feature generation for asset classification. An embodiment includes generating an input document that includes a plurality of text fields of attribute data associated with a group of cloud computing assets having a specified label for a specified class. In [0007]: extracting a set of candidate features from the attribute data using a feature extraction module that evaluates the attribute data using first and second machine learning models, wherein the first machine learning model scores terms in the input document and outputs a first portion of the set of candidate features ranked according to the term scores - the learning apparatus including: at least one processor for training, wherein the processor for training is configured to: use each document data for training included in a document data group for training as input of the machine learning model to acquire output data which is output for each document data for training; In [0007]: the embodiment also includes generating tag information for a remote computing asset using a machine learning classifier that predicts the tag information based on the feature-selection values. Other embodiments of this aspect include corresponding computer systems , apparatus, and computer programs recorded on one or more computer storage devices, each configured to perform the actions of the embodiment. In [0015] The computer system includes a processor, a computer-readable memory, and a computer-readable storage medium, and program instructions stored on the storage medium for execution by the processor via the memory. In [0007]: The illustrative embodiments provide for feature generation for asset classification. An embodiment includes generating an input document that includes a plurality of text fields of attribute data associated with a group of cloud computing assets having a specified label for a specified class. In [0007]: extracting a set of candidate features from the attribute data using a feature extraction module that evaluates the attribute data using first and second machine learning models, wherein the first machine learning model scores terms in the input document and outputs a first portion of the set of candidate features ranked according to the term scores in [0038]: In some embodiments, a feature extraction module includes a Deep Learning classifier module that includes a deep learning model for capturing both keywords and sequential patterns. In some such embodiments, the feature extraction module train s the deep learning model directly with values in each text field. Odibat does not explicitly disclose: - calculate, for a part of the document data for training from the document data group for training, a loss function representing a degree of difference between correct answer data and the output data for each document data for training based on the output data obtained for each document data for training and the correct answer data; and update the machine learning model based on the loss function. However, Londer discloses: - calculate, for a part of the document data for training from the document data group for training, a loss function representing a degree of difference between correct answer data and the output data for each document data for training based on the output data obtained for each document data for training and the correct answer data; In [0045]: the retriever engine 205 may be configured to update the query encoding model and the document encoding model using a match loss function that generates a match loss function result based on the query and the matching sub-documents. A loss function may be described generally as a function that computes the distance between the current output of an operation and the expected output . In accordance with various embodiments of the inventive concept, the reader loss function and/or the match loss function may comprise a regression loss function or a classification loss function. (BRI: the distance is the degree of similarity) - and update the machine learning model based on the loss function. In [0034]: the query encoding model and the document encoding model may be updated using a match loss function result between the query and the matching sub-documents . In [0045]: reader loss function may be applied to identify the matching sub-document having the lowest reader loss function result. This identified matching sub-document may then be associated with a truth label for the query and the AI reader engine 230 may be updated based on the reader loss function results associated with the matching sub-documents It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Odibat and Londer. Odibat teaches machine learning based document processing providing an evaluation value for the plurality of the documents. Londer teaches loss function. One of ordinary skill would have motivation to combine Odibat and Londer that can improve the accuracy of the inference or answer generated ( Londer [0047]) . In regard to claim 19 : Odibat discloses: - A non-transitory computer-readable medium storing an information processing program that is executable by the processor included in the information processing device to perform the information processing method according to claim 18. In [0111] Conclusion Any inquiry concerning this communication or earlier communications from the examiner should be directed to TIRUMALE KRISHNASWAMY RAMESH whose telephone number is (571)272-4605. The examiner can normally be reached by phone. 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, Li B Zhen can be reached on phone (571-272-3768). 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. /TIRUMALE K RAMESH/Examiner, Art Unit 2121 /Li B. Zhen/Supervisory Patent Examiner, Art Unit 2121 Application/Control Number: 18/458,135 Page 2 Art Unit: 2121 Application/Control Number: 18/458,135 Page 3 Art Unit: 2121 Application/Control Number: 18/458,135 Page 4 Art Unit: 2121 Application/Control Number: 18/458,135 Page 5 Art Unit: 2121 Application/Control Number: 18/458,135 Page 6 Art Unit: 2121 Application/Control Number: 18/458,135 Page 7 Art Unit: 2121 Application/Control Number: 18/458,135 Page 8 Art Unit: 2121 Application/Control Number: 18/458,135 Page 9 Art Unit: 2121 Application/Control Number: 18/458,135 Page 10 Art Unit: 2121 Application/Control Number: 18/458,135 Page 11 Art Unit: 2121 Application/Control Number: 18/458,135 Page 12 Art Unit: 2121 Application/Control Number: 18/458,135 Page 13 Art Unit: 2121 Application/Control Number: 18/458,135 Page 14 Art Unit: 2121 Application/Control Number: 18/458,135 Page 15 Art Unit: 2121 Application/Control Number: 18/458,135 Page 16 Art Unit: 2121 Application/Control Number: 18/458,135 Page 17 Art Unit: 2121 Application/Control Number: 18/458,135 Page 18 Art Unit: 2121 Application/Control Number: 18/458,135 Page 19 Art Unit: 2121 Application/Control Number: 18/458,135 Page 20 Art Unit: 2121 Application/Control Number: 18/458,135 Page 21 Art Unit: 2121 Application/Control Number: 18/458,135 Page 22 Art Unit: 2121 Application/Control Number: 18/458,135 Page 23 Art Unit: 2121 Application/Control Number: 18/458,135 Page 24 Art Unit: 2121 Application/Control Number: 18/458,135 Page 25 Art Unit: 2121 Application/Control Number: 18/458,135 Page 26 Art Unit: 2121 Application/Control Number: 18/458,135 Page 27 Art Unit: 2121 Application/Control Number: 18/458,135 Page 28 Art Unit: 2121 Application/Control Number: 18/458,135 Page 29 Art Unit: 2121 Application/Control Number: 18/458,135 Page 30 Art Unit: 2121 Application/Control Number: 18/458,135 Page 31 Art Unit: 2121 Application/Control Number: 18/458,135 Page 32 Art Unit: 2121 Application/Control Number: 18/458,135 Page 33 Art Unit: 2121 Application/Control Number: 18/458,135 Page 34 Art Unit: 2121 Application/Control Number: 18/458,135 Page 35 Art Unit: 2121 Application/Control Number: 18/458,135 Page 36 Art Unit: 2121 Application/Control Number: 18/458,135 Page 37 Art Unit: 2121 Application/Control Number: 18/458,135 Page 38 Art Unit: 2121 Application/Control Number: 18/458,135 Page 39 Art Unit: 2121 Application/Control Number: 18/458,135 Page 40 Art Unit: 2121 Application/Control Number: 18/458,135 Page 41 Art Unit: 2121 Application/Control Number: 18/458,135 Page 42 Art Unit: 2121 Application/Control Number: 18/458,135 Page 43 Art Unit: 2121 Application/Control Number: 18/458,135 Page 44 Art Unit: 2121 Application/Control Number: 18/458,135 Page 45 Art Unit: 2121 Application/Control Number: 18/458,135 Page 46 Art Unit: 2121
Prosecution Insights