# US 12,008,305 B2 · Learning device, extraction device, and learning method for tagging description portions in a document > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 12,008,305 B2 - Document ID: US12008305B2 - Status: Granted - Kind: B2 - Application number: 17/272,700 - Filed: 2019-09-04 - Published: 2021-11-04 - Granted: 2024-06-11 - Earliest claimed: 2018-09-19 - Assignees: Nippon Telegraph and Telephone Corporation - Applicants: Nippon Telegraph and Telephone Corporation - Inventors: Takeshi Yamada - Web page: https://patentgrep.com/patent/US12008305B2 ## CPC - G06F 40/117: Tagging; Marking up (details of markup languages G06F40/143); Designating a block; Setting of attributes (style sheets, e.g. eXtensible Stylesheet Language Transformation [XSLT], G06F40/154) - G06F 40/216: using statistical methods - G06F 40/279: Recognition of textual entities - G06N 7/01: Probabilistic graphical models, e.g. probabilistic networks - G06N 20/00: Machine learning - G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N) - G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS ## Abstract An extraction apparatus includes: a pre-processing unit configured to perform, on training data that is data described in natural language and in which a tag has been provided to an important description portion in advance, pre-processing for calculating an information gain that indicates a degree of relevance to the tag for each word and deleting a description portion with low relevance to the tag from the training data based on the information gain of each word; and a learning unit configured to learn the pre-processed training data and generate a list of conditional probabilities relating to the tagged description portion. ## Claims ### Claim 1 A learning apparatus comprising one or more processors configured to perform operations comprising: obtaining training data that includes a plurality of description portions of a natural language document and a plurality of tags, wherein each description portion includes a set of words and each description portion is associated with a tag that was provided to the description portion in advance; calculating an information gain for each word included in the training data that indicates a degree of relevance of the word to the plurality of tags for the word; deleting, from the training data, words whose degrees of relevance do not satisfy a threshold to obtain pre-processed training data; learning the pre-processed training data and the information gain for each word that indicates the degree of relevance of the word to the plurality of tags; generating a list of conditional probabilities indicating a trend for attaching the plurality of description portions with the plurality of tags; receiving test data; tagging, based on the list of conditional probability, description content of the test data by attaching a tag, included in the plurality of tags, to a corresponding portion of the description content; extracting a test item from the test data based on the tagged description content; and outputting the test item into a testing apparatus for testing. ### Claim 2 (depends on 1) The learning apparatus according to claim 1, wherein the operations comprise: deleting a sentence that does not include a noun for which the information gain is higher than a predetermined threshold value from the training data. ### Claim 3 (depends on 1) The learning apparatus according to claim 1, wherein the operations comprise: deleting a sentence that does not include a verb but that includes a noun for which the information gain is higher than a predetermined threshold value from the training data. ### Claim 4 An extraction apparatus comprising one or more processors configured to perform operations comprising: obtaining training data that includes a plurality of description portions of a natural language document and a plurality of tags, wherein each description portion includes a set of words and each description portion is associated with a tag that was provided to the description portion in advance; calculating an information gain for each word included in the training data that indicates a degree of relevance of the word to the plurality of tags for the word; deleting, from the training data, words whose degrees of relevance do not satisfy a threshold to obtain pre-processed training data; learning the pre-processed training data and the information gain for each word that indicates the degree of relevance of the word to the plurality of tags; generating a list of conditional probabilities indicating a trend for attaching the plurality of description portions with the plurality of tags; receiving test data; tagging, based on the list of conditional probability, description content of the test data by attaching a tag, included in the plurality of tags, to a corresponding portion of the description content; extracting a test item from the test data based on the attached tags of the description content; and outputting the test item into a testing apparatus for testing. ### Claim 5 A learning method to be executed by a learning apparatus, the learning method comprising: obtaining training data that includes a plurality of description portions of a natural language document and a plurality of tags, wherein each description portion includes a set of words and each description portion is associated with a tag that was provided to the description portion in advance; calculating an information gain for each word included in the training data that indicates a degree of relevance of the word to the plurality of tags for the word; deleting, from the training data, words whose degrees of relevance do not satisfy a threshold to obtain pre-processed training data; learning the pre-processed training data and the information gain for each word that indicates the degree of relevance of the word to the plurality of tags; generating a list of conditional probabilities indicating a trend for attaching the plurality of description portions with the plurality of tags; receiving test data; tagging, based on the list of conditional probability, description content of the test data by attaching a tag, included in the plurality of tags, to a corresponding portion of the description content; extracting a test item from the test data based on the tagged description content; and outputting the test item into a testing apparatus for testing.