Entity-Augmented Distributional Semantics for Discourse Relations

摘要: Discourse relations bind smaller linguistic elements into coherent texts. However, automatically identifying discourse is difficult, because it requires understanding the semantics of linked sentences. A more subtle challenge that not enough to represent meaning each sentence a relation, relation may depend on links between lower-level elements, such as entity mentions. Our solution computes distributional representations by composition up syntactic parse tree. key difference from previous work compositional we also compute for mentions, using novel downward pass. are predicted only sentences, but their coreferent The resulting system obtains substantial improvements over state-of-the-art in predicting implicit Penn Treebank.