New Method of Character Relations Auto-Extraction

A new method of character relations auto-extraction from Chinese text set is proposed in this paper. In this method, characters and their relations are extracted from sentences and documents by computing the relative strength and attributive weight of relative properties. Each dispersed relation is organized and represented by the relative coefficient. For the purpose of exhibiting relations visually, the relations map, which describes characters and relative properties between them, could be constructed based on the coefficients. The experiment of relation extraction shows that the precision and recall of this method are 85.56% and 61.94%; the experiment of relations map shows that the precision and recall are 94.74% and 52.94% of level-1, 96.88% and 59.62% of level-2, respectively.