Detecting and Visualizing Coordinate Relationships over Time

摘要

テキスト情報から知識を発見する研究がこれまで多く行われrnてきた．Hearst［1］は，大量のテキストコーパスから上位下位rn関係を発見する研究を行い，そこでは，“such as”のような構rn文パターンが用いられている．近年，Web上のコンテンツが増rn大していることから，Webを情報源として知識発見を行う研究rnも行われている．%Mining search engines for different kinds of relationships between entities, such as coordinate relationships, has recently attracted much research interest. However, detecting the evolution of such relationships over time has not been researched well. Certain document repositories, such as news archives or blog collections often enable temporal search functionality that can deliver results with respect to selected temporal constraints. In this paper, we demonstrate a method for detecting changes in coordinate terms over time by mining a large news archive through its search interface. We present also a visualization system that lets users explore and understand the evolution of coordinate relationships between different entities.