This paper proposes a Chinese parsing algorithm based on the classified linguistic attribute knowledge and statistical disambiguation mechanism. The algorithm treats the Chinese paring procedure as a heuristic processing of selecting the preferred one from all the candidate syntactic trees, which includes two steps. The first is the construction of syntactic-tree candidate set by using classified linguistic attribute knowledge and GLR algorithm; the second is the best tree selection by using statistical disambiguation mechanism. We focus on the discussions of attribute knowledge classification and statistical decision-making in this paper. Based on this proposed methodology, the qualitative and quantitative knowledge is integrated interleavingly and cooperatively, not only the advantages of the two kind knowledge are kept, but also the burden of knowledge acquisition is reduced greatly, furthermore the robustness of the linguistic knowledge-based method is improved significantly by the classification.
展开▼
