Summary: | Web based search for a specific problem usually returns long lists of results, which may take up a lot of time to browse until finding the exact solution, if found at all. Community Question Answering systems on the other hand offer a good alternative to solve problems in a more efficient way, by directly asking the community, or automatically extract similar questions that have already been answered by other users. Using external knowledge bases for such similarity measures is a growing field of research, due to their rich content and semantic relations. Indeed, many research works base their semantic textual similarity measures on annotating texts or extracting specific knowledge from an external knowledge base.
Our research aims at creating a semantic domain-specific graph of keywords using data extracted from the DBpedia knowledge base. This keywords graph will be used later, in a graph-based similarity approach inside a CQA archive in order to retrieve similar questions. In this paper, we define the structure of the semantic graph and propose our method for automatically creating it, backed with experimental results.
|