| 摘要 |
PROBLEM TO BE SOLVED: To correctly estimate matching degree of a document to a retrieval word independently of types of the document. SOLUTION: This system computes a TF term obtained by reflecting frequency of an inputted retrieval word in a target document and an IDF term obtained by reflecting importance of the retrieval word based on information of a multi-document information storing means, and computes document matching degree indicating the degree of matching between the target document and the inputted one or a plurality of retrieval words from the TF term and the IDF term for the retrieval word. An expectation value of appearance frequency of the retrieval word t in the target document d when the document d is included in a document setσ(t) appropriate to the retrieval word t is calculated by approximating the document setσ(t) by a document setκ(t) being the whole documents where the retrieval word t appears, and the difference between the expectation value and the appearance frequency that the retrieval word t actually appears in the target document d is reflected in the TF term. COPYRIGHT: (C)2006,JPO&NCIPI
|
