Abstract
How to develop a conceptual structure to represent the knowledge encapsulated in text or multimodal presentation is a major challenge for semantic processing. This paper describes an approach based on a mapping process that develops such conceptual structure by projecting the verbal content onto a knowledge domain. However, for the target knowledge domain to support the mapping process, it needs to posses some specific features e.g. to be robust and flexible enough to organize and codify the contextual knowledge which guides the interpretation of presentation contents, and rich enough to incorporate knowledge in the presentation. To fulfil this requirement, the paper proposes an ontologybased knowledge model that satisfies the set of desirable properties. The model contains three layers: ontology layer, qualitative layer and quantitative layer, each of them is associated with the relevant kinds of knowledge. Ontology is developed to capture the conceptual understanding of the domain in which presentation takes place, and provide a semantic basis for grounding the fine-grained knowledge which is represented in the other knowledge layers. The qualitative layer offers significantly qualitative description of the domain, such as low-level feature of physical process and explicit causal process, while the quantitative layer is meant to capture the mathematical properties of the objects. A conceptual structure, generated as an outcome of the semantic processing, will retain the model's three-layered representation structure.