Partially lexicalized parsing model utilizing rich features

In this paper, we propose a partially lexicalized parsing model utilizing rich features to improve the parsing ability and reduce the parsing cost. In order to disambiguate parse trees effectively, it employs several useful features such as a syntactic label feature, a content feature, a functional feature, and a size feature. Besides, it is partially lexicalized so as to reduce the parsing cost closely connected with lexical information. Moreover, it is designed to be suitable for representing word order variation and constituent ellipsis in Korean sentences. Experimental results show that the proposed parsing model using more features performs better although it less depends on lexical information.