基于依存关系的语义角色标注

摘要

针对以句法成分为基本标注单元语义角色标注的瓶颈问题,描述一个以依存关系为标注单元的语义角色标注系统,经过依存关系分析、谓词标识、特征抽取、角色识别和角色分类,最终在CoNLL2008SRLSharedTask自动依存分析的WSJ测试集取得了较好的结果,F1值达到了80.94%,结果证明其性能明显好于基于句法分析的SRL.%Aiming at the bottlenecks of syntactic wee-based Semantic Role Labeling(SRL), this paper explores dependency relationship-based semantic role labeling. By properly integrating dependency parsing, predicate identification, feature extraction, semantic role identification and sernantic role classification, this system achieves the FI measure of 80.94% on the WSJ portion of the CoNLL2008 SRL Shared Task, using automatic dependency parsing. Experimental results show that it is better than the FI measure of syntactic tree-based semantic role labeling apparently.