Passage-based Answer Extraction for List Questions
Speaker: Fang Xu
Abstract:
My talk is on list question answering . Answer Extraction for list questions has traditionally followed the schema of factoid question answering and focused on acquiring knowledge about particular question-answering(QA) relationships within sentence. However, according to analysis on related documents, those relationships are not only restrained in one sentence. Many list questions can only be appropriately answered by learning paragraph- or document-level contextual properties. My motivation is to develop a novel extraction paradigm that automatically extract paragraphs and discover QA relations from them to determine final answers. My work is focusing on exploring new representations and methods for this task.