Research On History-based Chinese Hierarchical Parsing

With the increasingly power of computer capacity and the fast development of the Internet, the informativeness of the human society is becoming higher and higher. As an important carrier of information, computer processing of natural language is critical in the information society. This paper focuses on syntactic parsing, the fundamental problem in natural language processing.This paper follows history-based approaches, such as the one in (Ratnaparkhi 1999), and explores a hierarchical parsing strategy by constructing a parse tree level by level, which includes part-of-speech tagging, phrase chunking, and structural parsing. The intuition behind our strategy is that simple constituents should be constructed first so that the complex ones can rely on richer contextual information in the following passes. This is done as follows: given a forest of trees (especially at beginning, each word is regarded as a single tree), we recursively recognize simple constituents first and then form a new forest with a less number of trees until there is only one tree in the newly produced forest.In addition, this paper also integrates Chinese word segmentation into the hierarcical parsing strategy.Evaluation on the the Chinese Penn Treebank shows that our hierarchical parsing strategy works well on the Chinese language and achieves comparable performance with the state-of-the-art ones.