Not a simple yes or no: Uncertainty in indirect answers

Marie-Catherine de Marneffe, Scott Grimm and Christopher Potts

SIGDIAL Workshop on Discourse and Dialogue (SIGDIAL 2009)
Queen Mary University of London, September 11-12, 2009


Dialogues contain responses to polar questions which are neither `yes' nor `no', but are indirect responses providing information which serves to resolve the query. This paper presents the first treatment of such indirect question-answer pairs (IQAPs). We first present empirical results from a corpus analysis of indirect responses observed in natural dialogues, demonstrating that they are a pervasive discourse technique, upon which we establish a classification of indirect answers, cross-classified by the degree to which they resolve the question under discussion and the type of information used in the relevant inference. We then model the different types of IQAPs using Markov logic networks, which allow rich inferential reasoning on relations by combining the power of first-order logic and probabilities to cope with uncertainty. This method permits to represent uncertainty associated with indirect answers that only partially-resolve a query and leads to a robust model of question-answer resolution in dialogue.