Departmental Bulletin Paper 潜在意味に基づく新聞記事からの特徴抽出

奥村, 直也

In recent years, we can get huge documents from the internet. These documents havehuge amount of information. Very often, in news articles, headlines contain characteristic expressions speci c to their contents. In this investigation, we propose new methods to generate headlines for news articles with automatic estimations. Here we examine both news articles and the headlines separetely, bridge two documents in such a way that similar articles have similar headlines. In experiments, we show the effectivi ties of our approaches by some experimental results.Key Words : News Articles, Latent Semantics, Headline Generation, Dependency Structure,Feature Extraction

Number of accesses :  

Other information