We propose new methods to make literature abstracts. In documents written by Japanese, subject is often omitted in a sentence. Therefore a sentence without Subject is measured as an important and selected as an abstract, it is difficult for us to read. And the abstracts are evaluated subjectively by human. This is the reason that we need the standard to evaluate the abstracts objectively. To solve these problems, we take the approach as following. At first, we identify Subject in an each sentence and analysis in anaphora to restore Subject. Secondly, we measure each sentence based on our proposed weighting. Finally, we extract the sentence, by its weight, as an abstract. As an objective evaluation standard, we propose the search precision. In this paper, we applied this method to news articles browsed in web and make their abstracts. As results, the abstracts meet our need and we evaluate them by the search precision.
展开▼