Discourse parsing for multi-party chat dialogues

摘要: In this paper we present the first ever, to best of our knowledge, discourse parser for multi-party chat dialogues. Discourse in dialogues dramatically differs from monologues since threaded conversations are commonplace rendering prediction structure compelling. Moreover, fact that data come chats renders use syntactic and lexical information useless people take great liberties expressing themselves lexically syntactically. We dependency parsing paradigm as has been done past (Muller et al., 2012; Li 2014). learn local probability distributions then MST decoding. achieve 0.680 F1 on unlabelled structures 0.516 fully labeled which is better than many state art systems monologues, despite inherent difficulties have.