Abstract
We propose a novel approach to the problem of expertise mining in an enterprise, taking advantage of online social applications deployed within the enterprise. Based on the assumption that the users' interactions with such social software reflect to some extent their expertise, we devise a probabilistic method for identifying the main areas of expertise of users based solely on their set of tags extracted from a social bookmarking system. We base our approach on statistical language models, which we adapt to fit our unique setting. We train and validate our model on a real world dataset extracted from two IBM-internal applications. Our results show that our approach provides a viable alternative to other methods that rely on documents extracted from the enterprise corpora. © 2009 IEEE.