iMERGE : Interactive Ontology Merging

In this paper we present novel visual analytics techniques which help the user in the process of interactive ontology mapping and merging. A major contribution will be the strong integration and coupling of interactive visualizations with the merging process enabling the user to follow why concepts are merged and at which position in the ontology they are merged. For this purpose, adapted ontology similarity measures and new techniques for representing ontologies will be required to enable responsive, real time visualization and exploration of the comparing and merging results. 1 Visual Ontology Mapping The areas of ontology mapping and ontology merging have largely relied on automatic and semi-automatic methods in the past (FOAM [1], PROMPT [2], OLA [3] and FCA-Merge [4]), where user control and interaction is limited and results are typically only presented to the user at the end of some complex computational process. The effectiveness of these approaches can be increased in many application domains, if these approaches use the users knowledge and expertise in the comparing and merging process and support the explorative and iterative activities that are essential for the user’s sensemaking process. Ontology merging is still largely a human-mediated process. The user could not trust in automatic results, where he does not know where and for which reason concepts are merged. In this paper we present an approach, which helps to enable users to explore the ontology and to compare results in an intuitive and efficient manner. We aim to support the analytic comparing and merging process providing tightly linked and integrated techniques and views for visualizing and exploring the raw ontologies and derived merging results. For this purpose we develop a prototype editor iMerge. The introduced editor iMerge provides different views for this purpose(see Fig. 1). The SmartTree-View [5] extends the conventional tree widget with a number of mechanisms facilitating ontology exploration and development. In addition to the hierarchical structure shown (typically the class hierarchy), non-hierarchical relations are shown dynamically upon selecting a node. SmartTree introduces Condense+Expand and Prune+Grow, two new interaction techniques allowing to hide and expose parts of the tree. The Matrix-View [6] is suited for comparing two ontologies and determining where most of the mappings between ontologies occur. In the Matrix-View the