DOCUMENT RETRIEVAL SYSTEM

摘要

PROBLEM TO BE SOLVED: To improve the precision and efficiently of document retrieval based upon a non-complete-matching model by retrieving a document including a character string which is not registered in a dictionary without any omission and efficiently calculating the similarity between the document and a question sentence even if a user questions with a question sentence including the character string which has not been registered in the dictionary. SOLUTION: A similarity decision means 1605 calculates the similarity between each document and an input intention by using statistic information on words previously gathered in a word statistic information storage means 1604 when the input character string from the user is a word or by dynamically generating statistic information by using a full-text retrieving means 1606 when the input character string is not a word. A dictionary modifying means 1611 properly modifies the dictionary 1602 by using information stored in an input history storage means 1608 and a language information storage means 1610 and then a word statistic information gathering means 1603 gathers word statistic information again.