紀要論文

新聞記事群からのイベント抽出と時間情報に基づく編成

Organizing Events Extracted from Multiple Document by Time

jpn

18810667

https://doi.org/10.15002/00010525

法政大学大学院情報科学研究科

9

77

82

2014-03

Version of Record

When the news such as a large-scale or long-term case, accident or disaster is reported, they take a long time to arrange events which happened in these cases by human because they have huge data. This paper suggests a method for quickly picking up information about events including when and what accidents happened and arranging them when the case occurred. To pick up information from news articles, we use named entity extraction and an original rule which helps picking up particular extraction. Such a rule can extract more information from news articles than using only named entity extraction. In addition, as well as comparison of the time when events happened, events are united which report same or similar events by sentence from news articles for users of this system to check easily when the flow of events is visualized. To confirm the precision of the method, this paper checked how results reproduced real events and what informations are changed by addition of new events from collected news.

Article

AA12222297