StoryTracker: A Semantic-Oriented Tool for Automatic Tracking Events by Web Documents

Media vehicles play an essential role in investigating events and keeping the public informed. Indirectly, logs of daily events made by newspapers and magazines have been built rich collections of data that can be used by lots of professionals such as economists, historians, and political scientists. However, exploring these logs with traditional search engines has become impractical for more demanding users. In this paper, we propose Story Tracker, a temporal exploration tool that helps users query news collections. We focus our efforts (i) to allow users to make queries by adding information from documents represented by word embbedings and (ii) to develop a strategy for retrieving temporal information to generate timelines and present them using a suitable interface for temporal exploration. We evaluated our solution using a real database of articles from a huge Brazilian newspaper and showed that our tool can trace different timelines, covering different subtopics of the same theme.