Title: Reconciling Coreference Resolution with Traditional Knowledge Sources Abstract: The resolution of coreference relations requires many knowledge sources including surface properties, syntactic, semantic information, and word knowledge. Early coreference resolvers use various information sources and complex heuristics including syntactic and semantic parallelism, gender and number agreements, centering theory and estimated word knowledge. However, there is no trace of an explicit use of the above features in recent coreference resolution systems. Instead, lexical features are now the major source of information. In this talk, we introduce a possible solution to bring back the use of traditional knowledge sources in state-of-the-art coreference resolvers. We achieve significant improvements in both in-domain and out-of-domain evaluations.