Abstract
In this paper we introduce a method for searching appropriate articles from knowledge bases (e.g. Wikipedia) for a given query and its context. In our approach, this problem is transformed into a multi-class classification of candidate articles. The method involves automatically augmenting small knowledge bases using larger knowledge bases and learning to choose adequate articles based on hyperlink similarity between article and context. At run-time, keyphrases in given context are extracted and the sense ambiguity of query term is resolved by computing similarity of keyphrases between context and candidate articles. Evaluation shows that the method significantly outperforms the strong baseline of assigning most frequent articles to the query terms. Our method effectively determines adequate articles for given query-context pairs, suggesting the possibility of using our methods in context-aware search engines.