Using Mined Coreference Chains as a Resource for a Semantic Task

•Motivation: use output of coreference resolution system as a resource for semantic tasks •Coreference chains: complementary properties compared to other resources, such as cooccurrence statistics , e.g.: “cows” “cattle” vs. “cows” “milk” •Coreference-based similarity can be used as an additional feature for any task that distributional similarity is useful for (e.g. finding alternative names for entities, knowledge base population) • Task here: detecting antonyms ⇒ Antonyms: distributionally similar but semantically dissimilar words ⇒Distributional models often cannot distinguish them from synonyms