Semantic Role Labelling For Special Chinese Clauses

Language analyzing commonly comprise three levels : syntax, semantic, pragmatics. In the field of Natural Language Processing, one important way of acquiring semantic information is Semantic Role labeling (SRL). Semantic role labeling is a feasible proposal to shallow semantic parsing, which is labeling semantic role for natural language phrase for a given predicate verb and the natural language phrase as part of the framework of verb must be given the semantic meaning. This paper's research content is to make the current semantic role labeling system perfect, and can label the semantic role for every Chinese clause.For Chinese simple clause, we have built a semi-supervised system containing three steps for SRL. Firstly, fixing on the sentence's main verb and corresponding slots. Secondly, confirming the sentence's candidate semantic role frames. Lastly, selecting the right semantic role from the candidate set. The SRL system achieves the accuracy of 83.32%.According to the characteristics of two special Chinese clauses—ba-construction and bei-construction, a rule-based method of semantic role labeling is contrived. After the classification, chunking, semantic role labeling, the labeling accuracy of ba-construction and bei-construction reaches 88% and 93% respectively.Base on the syntax of the other verb in the shi(是)-construction, classifying the sentences into three classes. Employing the Chinese clause SRL system is for the first class. Using the rule-based method is for the second class and the third class.For the shi(使)-construction SRL, according to the driving force and result , splitting one sentence into two clauses . Split the causer + shi (使)+ causee + result into causer + shi (使) + causee and causee + result, make one sentence into two sentences, subsequently label the semantic role . The accuracy reaches 87%.