An Expert Experience Probabilistic Model for Enterprise Expert Finding

Finding experts accurately and automatically is becoming difficult especially in a large organization. This paper presents a probabilistic model which applies language modeling techniques to find experts in enterprise corpora. The expertise of each candidate expert is modeled through the associated experience. We employ a qualification of experience, and validate this qualification as a measure of expertise to replace the documents. In our model, the relevance of a candidate's expertise to a given topic is estimated by its likelihood of generating topic words. Our evaluation on TREC test collection indicates the effectiveness of the probabilistic model in expert finding. This model is able to be improved by clustering of the relevant experience measures, which can also save retrieval time.