- 著者
-
蓮井 洋志
- 出版者
- 室蘭工業大学
- 雑誌
- 室蘭工業大学紀要 (ISSN:13442708)
- 巻号頁・発行日
- vol.50, pp.167-174, 2000-11-30
I have studied the tf.idf method as the automatic extraction of abstract in order to assist in retrieving the document database. I defined sentence importance as the sum of the word importance of tf.idf value, and the system extracted the several sentences which are the most important. In this paper, I propose the extended tf.idf method in order to get more comprehending abstract. It is the method which adds three ideas to the tf.idf method, the case weight, the eliminataion of demonstrative words and the conjunctions, and the decision of the region for the extraction. Result of the experiment shows that the abstract of the extended method is more comprehensible than one of the tf.idf method.投稿論文