A System for Collecting Tweets Using Event-Based Structuring of Web Contents

We introduce a system that aims to improve understanding of an event that appears on news and microblogs. In order to identify an event that a user is interested in, a user gives a system a URL of browsing news page. A system has two kinds of agents. One is a news agent to extract relevant news articles and the other is a microblog agent to extract related microblogs. A news agent extracts news articles that represent the same event as a browsing news article. A microblog agent extracts related microblogs from news articles. A microblog agent uses news URL references and a similarity between a news article and a microblog. As the result of experiment, tweets that are collected using our proposed method were 300 times more than tweets using a primitive method. We confirmed that a user could obtain more related tweets using our proposed system.