摘要 |
<P>PROBLEM TO BE SOLVED: To efficiently reduce a semantic drift. <P>SOLUTION: This data extraction device is configured to generate an identification model by learning processing using the property of the pair of a positive example entity and its attribute and the property of the pair of a negative example entity and its attribute as learning data with a teacher, and to input the property of the pair of an object entity and its object attribute to the identification model, and to identify whether or not the object entity is the positive example entity. At this time, the data extraction device outputs the result of determination of whether or not the object attribute is a positive example, and receives the input of manually corrected content, and determines a positive example attribute by using the manually corrected content. <P>COPYRIGHT: (C)2012,JPO&INPIT |