| 摘要 |
PROBLEM TO BE SOLVED: To provide an accurate information extraction technique by providing and narrowing down reliability information for extracted results through storing a semantic attribute and relation information extracted from a document together with time information including preparation date of the document or the like. SOLUTION: Patterns of a description are checked against a relation expression pattern dictionary or a semantic attribute pattern dictionary one by one (S1). Checking of words starts from the unchecked position in the document (S2). If a word string matching with a pattern (S3) exists, it is extracted and stored in the specified sequence (S4). Besides, together with the stored information, the time-related information of the document (time related information such as a preparation date and a issuance date of the document) is stored (S5). When checking of all words in the document has been completed (S6), checking against the next pattern is performed in the same way (S1). The time information itself of the document is extracted from the bibliography information or the header information. |
