Abstract: | In natural language queries/commands to an intelligent multimodal system, a particular kind of ambiguity - source ambiguity - can arise between items in a visual display and entities in the domain being represented. We associate with each symbolic entity an attribute called its source, which indicates which model (screen or domain) the entity belongs to. Also, each referring expression is assigned two representation: the described entity sets and the intended refernts. The process of resolving source ambiguity then happens as a part of the referent finding process. This is work in progress, with many interesting aspects needing further developement.
|