Towards a Unified Knowledge-Based Approach to Modality Choice

This paper advances a unified knowledge-based approach to the process of choosing the most appropriate modality or combination of modalities in multimodal output generation. We propose a Modality Ontology (MO) that models the knowledge needed to support the two most fundamental processes determining modality choice – modality allocation (choosing the modality or set of modalities that can best support a particular type of information) and modality combination (selecting an optimal final combination of modalities). In the proposed ontology we model the main levels which collectively determine the characteristics of each modality and the specific relationships between different modalities that are important for multi-modal meaning making. This ontology aims to support the automatic selection of modalities and combinations of modalities that are suitable to convey the meaning of the intended message.