Chinese Syntactic Parsing Based On Linguistic Entity Relationship Model

The syntactic parsing is a very crucial task in the natural language processing (NLP),and it is important for NLP researches such as machine translation and automaticquestion-answering. The target of syntactic parsing is constructing the syntactic tree ofexpressing the grammatical relations of words. The task of syntactic parsing is establishingthe parsing model and algorithm for automatically deriving the parsing tree.Many grammar models of syntactic parsing often have some problems. For example,the phrase structure grammar (PSG) has over-generative capability, and it can notdisambiguate the syntactic ambiguity effectively. And the PSG lacks semantic descriptionsand constraint informations in the parsing precedure. In addition, many Chinese syntacticparsing models also have some problems and therefore need to be improved furthermore.At present, the syntactic parsing effect of Indo-European languages such as English is better,but the Chinese syntactic parsing effect is much lower relatively for the special grammarcharacteristic of Chinese. For this reason, we consider the Chinese grammaticalcharacteristic and try to propose a linguistic entity relationship model for Chinese syntacticparsing. The work in this paper is presented as follows.(1) A linguistic entity relationship model for Chinese syntactic parsing is proposed. Inthis model, linguistic entity relationship modes (RM) and their corresponding syntacticconstituent-function expressions (SCFE) are used as the foundation of syntactic parsing.Multivalue recursive function of words is used for syntactic parsing of chunks in SCFE.Preferential strategies comprise both syntactic and semantic weight strategies, and are usedfor selecting the optimal parsing tree. The model considers many Chinese grammaticalfeatures, such as Chinese sentence pattern, predicate-argument attribute of verb, chunks ofsentence. It can effectively overcome the problems in PSG, such as over-generativecapability and syntactic ambiguity.(2) The relationship modes (RM) and their corresponding syntacticconstituent-function expressions (SCFE) are proposed for describing the Chinese sentencepatterns. In the traditional PSG, the predicate can not express complex syntactic features. This leads to the problems of over-generative capability and syntactic ambiguity. Forovercoming the problems, RM and SCFE are proposed. RM and SCFE can guide thesyntactic parsing and therefore effectively decrease the syntactic ambiguity.(3) The multivalue recursive function of words is proposed for defining and parsingthe chunks in SCFE. It can describe the grammars and rules of the chunks.(4) The preferential strategies are proposed for selecting the optimal tree fromsyntactic forest of containing many valid parsing trees. In the preferential strategies, boththe syntactic and semantic features of the parsing trees are considered. Furthermore, theflexible semantic matching is used for avoiding some problems of rigid semantic matching.We develop a parser based on the model for the experiment. The experimental resultshows that the parser has better performance compared with other parsers, and that themodel is feasible and effective.