| 摘要 |
<P>PROBLEM TO BE SOLVED: To always secure high ranking accuracy, and to apply an arbitrary learning algorithm by using learning data relating to data to be searched in an information search device. <P>SOLUTION: In the document set of the same classification (target domain) which is the same as that of a search object, correct answer data in an original domain which is similar with the correct answer data in the target domain is selected by using a ranking function leant from data (correct answer data) in which the correct sequence of documents with respect to each query is already known and correct data in the document set of the classification (original domain) which is not homogeneous with but relevant to that of the search object, and a ranking function is estimated from the correct answer data in the target domain and the correct answer data in the original domain selected by a sample selection part, and the respective documents to be searched are sequenced with respect to a new query by using the estimated ranking function. <P>COPYRIGHT: (C)2012,JPO&INPIT |
