Guarded Parsing of Chinese Sentence without Word Segmentation

Abstract

One problem in Chinese language processing is the lack of grammatical regularity in the language. This leads to very complex Chinese grammar in order to obtain satisfactory results, which in term increases the complexity in the parsing process. In order to simplify this process, it is desired that simple grammar is to be used. However, this will cause ambiguities in the parse result. Word segmentation is an tradition preprocessor to a Chinese parser. Unfortunately, word segmentation is a process which is error-prone, and once an error is made, the parsing connot be recovered. In this paper, a preprocessor called lexical scanner is used to replace the word segmentation process. The lexical scanner is a integrated part of the parserm and is a kind of segment on the fly approach. However, this has to be combined with guarded parsing in order to achieve good results. In this paper, we introduced a technique called Guarded Parsing which was first introduced by us for syntactic pattern recognition [1]. In this approach, simple grammar is used so that the effort in building and maintainin gthe grammar is much reduced. The parsing process will also be simplified. Ambiguity is resolved by the use of guards which not only checks the syntactic compatibility, but also take semantic information into consideration. This work is a part of the Hong Kong University Machine Translation (HKUMT) project. The project is supported by the Industry Department, Hong Kong Government.