Chinese chunking and its application on similarity computation

This paper presents a new Chinese chunking algorithm based on conditional random fields. Conditional random fields overcome the label bias problem, model the labeling sequence and utilize many types of features. Furthermore, an algorithm of chunk similarity computation is proposed based on the systematic similarity method and semantic dictionary. The experimental results show that this approach achieves impressive accuracy in terms of the F-score: 92.00%. And the similarity computation algorithm performs well.