Semantic Role Labeling System For Chinese Clauses

Recent years have seen an increasingly interest in shallow semantic parsing, namely the semantic role labeling task (SRL). The lack of semantically hand-annotated corpus in Chinese and the complicated constructions of Chinese sentences arise as an important problem of SRL for Chinese. However, there have been few researches in Chinese semantic analysis to tackle these problems. In this paper, we will present a first overall result of a semi-supervised semantic role labeling system for Chinese clauses, using only a small set of labeled data and a large collection of unlabeled ones.Unlike the existed SRL systems which are straightly built for all the sentences in corpus, our system focuses on the simple clauses with only one verb phrase (1-VP clauses). Clauses which have more than one verb phrases (multi-VP clauses) have been transformed into 1-VP clause pattern and then are available for the system. The ability of our SRL system dealing with simple clauses is considered as the basic performance of the system. When processing more complex clauses (namely clauses with more than one verb phrase in our definition of complex clauses), it is difficult to identify the main-predicate and the larger functional chunks. The final accuracy of semantic role labeling decreases badly due to the difficulty. We propose in this paper a novel way to replace the auto chunker for clauses with more than one verb phrase (multi-VP). The multi-VP clauses will be transformed to 1-VP forms which could be handled easily by our SRL system. It will be proven that the multi-VP clauses can be labeled in an almost same accuracy as the basic accuracy for 1-VP clauses after being transformed.Our work in building such a complete semi-supervised SRL system will be helpful for further semantically study in Chinese. Detailed descriptions and analytic experimental results of our semi-supervised SRL system will be displayed in this paper.