Using Semantic Roles for Coreference Resolution

In this paper, we systematically explore the use of semantic roles in coreference resolution. Here, the semantic roles are automatically determined using a state-of-the-art SRL system, ASSERT, and integrated into the resolution by combining them with detailed pronoun categories. In particular, the anaphoricity of the pronoun "it" in "it" sentence pattern is determined using its semantic role and applied in a filter process. Experimental results on the ACE 2003 corpus show that our method can significantly outperform the baseline system by 3.1% in F-measure.