Coreference resolution systems can benefit greatly from inclusion of global context, and a number of recent approaches have demonstrated improvements when precomputing an alignment to external knowledge sources. However, since alignment itself is a challenging task and is often noisy, existing systems either align conservatively, resulting in very few links, or combine the attributes of multiple candidates, leading to a conflation of entities. Our approach instead performs joint inference between within-document coreference and entity linking, maintaining ranked lists of candidate entities that are dynamically merged and reranked during inference. Further, we incorporate a large set of surface string variations for each entity by using anchor texts from the web that link to the entity. These forms of global context enables our system to improve classifier-based coreference by 1.09 B~3 F1 points, and improve over the previous state-of-art by 0.41 points, thus introducing a new state-of-art result on the ACE 2004 data.
展开▼