Semantic Role Labelling For Complicated Chinese Clauses

In the field of natural language understanding, semantic role labeling is an important means to obtain semantic information,for the predicate verb argument and ancillary components of the sentence marking him as a semantic role, such as the causal agent, patient, time and location, etc.,. Assigning semantic roles for predicate verb argument and ancillary components of the sentence, can find the shallow semantic structure of sentences.In this paper, mainly to fill the part of the gap in existing Chinese semantic role labeling system that semantic role labeling for complicated clauses on the 2VP, so that the system for any English sentence can effectively semantic role labeling.In the Chinese language, the proportion of complicated clauses containing two or more verb accounted for more than a quarter of all sentences, but there will be no more satisfied with the current Chinese parser that can analysis syntactic tree completely for this type of Chinese clause.The errors identification in main verb and its associated argument structure will be an enormous deterioration for the final performance in semantic role labeling system. In the Chinese language processing, and no other method can well solve the problem.With the targeted researching, this article proposed a method that analysis the sentence argument structural using the characteristics extracted by the shallow POS tagging information only, based on a complete syntactic analysis can not be obtained by circumstances,so that arrive at the sentence blocks of information, and then use the rules for semantic role labeling.The method proposed is more stringent and more reasonable compared with the previous sentence classification criteria, by the comparing test, the accuracy rate of sentence classification obviously improved than the past. Since the adoption of the classification taking a better mapping semantic role as the principle, the performance of semantic role labeling has been greatly improved, while the method also make up the shortcomings that the entire semantic role labeling system can not recognize some particular sentence,and the performance of the overall semantic role labeling system is improved.