Extraction of Context Information from Web Content Using Entity Linking
平田紀史, 白松俊, 大囿忠親, 新谷虎松


We developed a system for extracting context information from microblogs, such as Twitter, using entity linking. A feature of Twitter is real-time. Our proposed system uses news articles to generate entity links since tweets are posted in real-time. It is difficult to extract context information from Twitter because the maximum length of a tweet is 140 characters. Therefore, we use both news articles and microblogs for entity linking. Entity links contain context entities. When our system extracts regional context, it uses entity links about geographical regions. The experimental results suggest that our proposed system can extract context entities based on Twitter users and news articles.