Paper: Optimising Incremental Generation for Spoken Dialogue Systems: Reducing the Need for Fillers

ACL ID W12-1509
Title Optimising Incremental Generation for Spoken Dialogue Systems: Reducing the Need for Fillers
Venue International Conference on Natural Language Generation
Session Main Conference
Year 2012
Authors

Recent studies have shown that incremental systems are perceived as more reactive, nat- ural, and easier to use than non-incremental systems. However, previous work on incre- mental NLG has not employed recent ad- vances in statistical optimisation using ma- chine learning. This paper combines the two approaches, showing how the update, revoke and purge operations typically used in in- cremental approaches can be implemented as state transitions in a Markov Decision Process. We design a model of incremental NLG that generates output based on micro-turn inter- pretations of the user?s utterances and is able to optimise its decisions using statistical ma- chine learning. We present a proof-of-concept study in the domain of Information Presen- tation (IP), where a learning agent faces the tra...