Entity-Relation Extraction for Chinese Based on Pattern Evolution and PageRank

Based on DIPRE, we use such knowledge as synonymy relation and hyponymy relation to evolve the extraction pattern in order to make the extraction pattern set become more expressive. Additionally, we use PageRank algorithm for reference to rank these entity-pairs in final result set so that the result set is global optimum. Experiments reveal that the new algorithm may not only enhance recall rate but greatly enhance precision rate.