| A probabilistic approach to the generation of speech act conditionals as indirect answers |
| Paper ID : 1063-ICIL |
| Authors |
|
Maryam Mohammadi * mohammadi@linguistics.rub.de |
| Abstract |
| This study aims at the discourse sensitive generation of speech act conditionals (SACs) as indirect answers in an interactive question answering system. SACs as indirect answers to polar questions do not only provide surplus information concerning the question, but also an epistemic indication to the relevance of the propositions. Our model is based on a probabilistic approach (Bayesian theory) to content determination that generates SACs by estimating the decision problem underlying the questions, while we assume that the speaker is blind to the user's requirements. Acceptability studies show that positive, negative and alternative SACs are appropriate structure for indirect answers in non-final question/answer sequences in a real estate domain where users ask about properties of apartments they take interest in. This paper focuses on the empirical studies of the project on Bayesian methods for preference-based content determination in a dialogue system (Stevens, 2016) that have been performed by the author. |
| Keywords |
| Speech act conditionals, probability, indirect answer, pragmatic |
| Status: Accepted |