Making Sense of Users’ Mouse Clicks: Abductive Reasoning and Conversational Dialogue Modeling

Intelligent information systems are expected to interpret the users' information needs semantically, taking the dialogue context into account. Whereas many research prototypes attempt to address the semantic interpretation of queries, only a few try to reason about other aspects of the user's individual dialogue behavior. This paper introduces an approach to context-dependent interpretation of ambiguous user dialogue acts in information seeking interactions. We illustrate the dialogue analysis and planning methods in the framework of the logic-based information retrieval system MIRACLE. Based on a dialogue model which describes potential developments of the interaction and recommended problem-solving steps, the abductive dialogue component (ADC) deals with unexpected user inputs which are ambiguous with respect to the intended course of action. Exploiting the dialogue history, the ADC uses abduction to generate interpretations of these inputs and thus to offer the user situation-dependent options for proceeding in the retrieval dialogue.