Incrementality all the way up

Abstract:

Dialogue is known to be incremental, with people able to make interpretations based on incomplete or partial information, and this underspecification is rife at all levels including sound, structure and meaning. We argue that underspecification and incrementality is also inherent in how people reason in dialogue, using enthymemes. We discuss how incremental reasoning can be elucidated in dialogue at points of mismatch, and provide the outline of a formal analysis using Type Theory with Records for one such example. This approach is compatible with existing dialogue models, and we believe that adding incremental reasoning abilities to conversational agents will allow them to be both more flexible and human-like in dialogue.

Research areas:

• Dialogue
• Reasoning
• IncReD

Year:

2017

Type of Publication:

In Proceedings

Book title:

Computing Natural Language Inference Workshop at the International Conference on Computational Semantics (IWCS)