Inference for Textual Question Answering
Papers from the AAAI Workshop
Sanda Harabagiu, Dan Moldovan, Srini Narayanan, Christopher Manning, Daniel Bobrow, and Ken Forbus Cochairs
Research in textual question answering has made substantial advances in the past few years. The state-of-the art in answering textual questions is beyond keyword matching. Processing of complex questions requires multiple forms of inference, for example abductions, default reasoning, inference with epistemic logic or description logic. Additionally, there are also forms of inference that are typical to language interpretation, for example conversational implicatures, processing of metonymies and metaphors. Often, the answer to questions involves temporal and spatial reasoning.
The challenge addressed by this workshop is posed by the identification, discussion and comparison of different inference mechanisms that operate on knowledge structures automatically derived from questions or candidate answers. Unlike inference schemes devised for manually-crafted knowledge, inference methods for question answering need to be robust, cover all ambiguities of language and operate on pragmatic information extracted from textual data. An important component of the workshop will be the discussion of available knowledge sources that can be used for inference of textual answers. This workshop constituted an occasion of bringing together researchers from the knowledge representation and reasoning (KRR) community with researchers that work in natural language processing (NLP).