Dedalo: Looking for clusters explanations in a labyrinth of linked data

We present Dedalo, a framework which is able to exploit Linked Data to generate explanations for clusters. In general, any result of a Knowledge Discovery process, including clusters, is interpreted by human experts who use their background knowledge to explain them. However, for someone without such expert knowledge, those results may be difficult to understand. Obtaining a complete and satisfactory explanation becomes a laborious and time-consuming process, involving expertise in possibly different domains. Having said so, not only does the Web of Data contain vast amounts of such background knowledge, but it also natively connects those domains. While the efforts put in the interpretation process can be reduced with the support of Linked Data, how to automatically access the right piece of knowledge in such a big space remains an issue. Dedalo is a framework that dynamically traverses Linked Data to find commonalities that form explanations for items of a cluster. We have developed different strategies (or heuristics) to guide this traversal, reducing the time to get the best explanation. In our experiments, we compare those strategies and demonstrate that Dedalo finds relevant and sophisticated Linked Data explanations from different areas.