Intra-chunk relationship analyse for Chinese base chunk labeling systems
Qiang Zhou · 2009
Responding to the flaws that most of the current base chunk parsers focus on the recognition of chunk boundaries and syntactic tags,this paper presents Chinese base chunking scheme which gives both the external syntactic description and the internal relationship in a base chunk.Firstly,an intra-chunk relation labeling set is used to describe the syntactic dependencies between words inside a chunk.Then,a statistical model is used to label each word in a sentence with a relationship tag.Finally,a finite state automaton is used to generate the base chunks in the Chinese sentence.Experiments show that the account method has little performance loss when labeling relationship in a base chunk compared with labeling syntactic tags using the classical IOB-based model.