A Model for Robust Chinese Parser

Computational Linguistics and Chinese Language Processing vol.1, no.1, August 1996, pp. 183-204 ÓComputational Linguistics Society of R.O.C. A Model for Robust Chinese Parser Keh-Jiann Chen* Abstract The Chinese language has many special characteristics which are substantially different from western languages, causing conventional methods of language pro- cessing to fail on Chinese. For example, Chinese sentences are composed of strings of characters without word boundaries that are marked by spaces. Therefore, word segmentation and unknown word identification techniques must be used in order to identify words in Chinese. In addition, Chinese has very few inflectional or grammatical markers, making purely syntactic approaches to parsing almost impos- sible. Hence, a unified approach which involves both syntactic and semantic information must be used. Therefore, a lexical feature-based grammar formalism, called Information-based Case Grammar, is adopted for the parsing model proposed here. This grammar formalism stipulates that a lexical entry for a word contains both semantic and syntactic feature structures. By relaxing the constraints on lexical feature structures, even ill-formed input can be accepted, broadening the coverage of the grammar. A model of a priority controlled chart parser is proposed which, in conjunction with a mechanism of dynamic grammar extension, addresses the problems of: (1) syntactic ambiguities, (2) under-specification and limited coverage of grammars, and (3) ill-formed sentences. The model does this without causing inef- ficient parsing of sentences that do not require relaxation of constraints or dynamic extension of the grammar. Keywords: Chinese Parser, Robust Parser, Information-based Case Grammar, Branch-and-Bound Algorithm 1. Introduction One of the steps in language understanding is producing the syntactic and semantic structure of a sentence. This step is called parsing. Other steps, such as inference, discourse analysis, and responding actions, can thus be performed accordingly. In general, sentence parsing is carried out by two different modules. The first module is a lexical analyzer which identifies the words in a sentence and provides their syntactic * Institute of Information Science, Academia Sinica,Taipei, Taiwan. E-mail: kchen@iis.sinica.edu.tw 183 and semantic information from a lexicon. From the output of the lexical analyzer, the second module produces the syntactic and semantic structure of the sentence. The syntactic structure and thematic role of each constituent are identified according to the grammar of the language. World knowledge is used in resolving structural ambiguities as well as in identifying thematic roles. It is well known that context-free languages, such as programming languages, can be easily parsed. Except for some very exceptional context-sensitive constructions, natural languages are context-free languages. In addition, if the depth of center-embedded constructions is bounded, natural languages are regular languages. That is to say, there is theoretically an efficient and simple parser for natural languages. Why then are there no natural language parsers that are practical and useful in general domain? Of course, there are parsers of limited success which can parse subsets of grammatical sentences in special domains. Stated simply, the difficulty is due to the complexity of the constraint relations between words, and not by the level of structural complexity. It …

Content truncated to 500 words