Efficient Representations for Multi-Modal Interaction

Contents

This position paper argues in favor of uniform, mixeddepth knowledge representation in intelligent systems that support flexible, multi-modal, mixed-initiative, two-way interaction. Uniformity has the advantage of minimizing knowledge interchange and can be implemented efficiently. Moreover, enhancements to the reasoner improve the performance of the entire system as all components use the same reasoner. Mixed-depth representations allow "shallow" or "deep" representations as required. Some representations can always be shallow, and this eliminates unnecessary interpretation. For example, mixed-depth representations permit the representation of ill-formed, vague, or ambiguous inputs, which can be further interpreted as necessary.