一个基于特征向量的语义角色标注

Semantic role labeling is a hot research in natural language processing. This paper described a feature-based semantic role labeling system. There are three steps in the process of semantic role labeling. It firstly identifies constituents that represent semantic arguments, then independently classifies constituents known to be semantic arguments into the specific categories, and finally in the post-processing step, deals with overlap arguments and constituents labeled with some core-arguments. Besides the features presented in other work, the paper extracts new features according to knowledge of grammar, pattern and collocation. The best SRL system achieves F1 value 77.54% and 78.75% on the development and WSJ test set respectively based on CoNLL-2005 Shared Task. As far as we know, it is the best result based on single syntactic parsers based on the corpus. The result also shows the effectiveness and robustness of the new extracted features.