Not a Simple Yes or No: Uncertainty in Indirect Answers

There is a long history of using logic to model the interpretation of indirect speech acts. Classical logical inference, however, is unable to deal with the combinations of disparate, conflicting, uncertain evidence that shape such speech acts in discourse. We propose to address this by combining logical inference with probabilistic methods. We focus on responses to polar questions with the following property: they are neither yes nor no, but they convey information that can be used to infer such an answer with some degree of confidence, though often not with enough confidence to count as resolving. We present a novel corpus study and associated typology that aims to situate these responses in the broader class of indirect question–answer pairs (IQAPs). We then model the different types of IQAPs using Markov logic networks, which combine first-order logic with probabilities, emphasizing the ways in which this approach allows us to model inferential uncertainty about both the context of utterance and intended meanings.