Conversational Semantic Role Labeling

Semantic role labeling (SRL) aims to extract the arguments for each predicate in an input sentence. Traditional SRL can fail analyze dialogues because it only works on every single sentence, while ellipsis and anaphora frequently occur dialogues. To address this problem, we propose conversational task, where argument be dialogue participants, a phrase history or current As existing datasets are sentence level, manually annotate semantic roles 3000 chit-chat (27198 sentences) boost research direction. Experiments show that traditional systems (even with help of coreference resolution rewriting) perform poorly analyzing dialogues, modeling histories participants greatly helps performance, indicating adapting conversations is very promising universal understanding. Our initial study by applying CSRL two mainstream tasks, response generation context rewriting, also confirms usefulness CSRL.