摘要 |
PROBLEM TO BE SOLVED: To automatically acquire summarized knowledge only by inputting an original and its summarized sentence without manually generating summarized knowledge needed for document summarization. SOLUTION: Word strings as differences between an original and its summarized sentence are extracted and information on word strings before and after a word string having a difference is obtained to automatically obtain summary knowledge. A morpheme analyzing device 1 performs a morpheme analysis of a pair of the original and its summarized sentence and outputs word divisions and parts of speech. A word correspondence device 2 calculates distances between words included in the original and its summarized sentence and calculates its optimum word correspondence on the basis of the distances between the words. A summarized word string extracting device 3 extracts words as differences from individual words of the optimum word correspondence, and when they are successive, they are put together into a word string, which is outputted so that the word strings of the original and those of the summarized sentence are made to correspond to each other. A summary condition extracting device 4 extracts and outputs word strings positioned before and after the word string of the original obtained as a difference. |