Abstract
This paper proposes an automatic digesting system for a Network Newsgroup “fj. wanted”. The key component of the system is the summary extraction from NetNews articles. In the summary extraction, 42 features are detected by using surface language expression patterns. Using these features, the system determines the category of the article and extracts the most important sentence (summary sentence) from the article. A blind test demonstrates that the accuracy of the category detection is 81% and the accuracy of the summary sentence extraction is 76%. The system creates the digest in HyperText Markup Language from the extracted summary sentences. The digest can be accessed via World-Wide Web.
