Association Concept Clustering Method Based on User-Generated Data

In order to overcome the inaccurate results in the traditional extraction and clustering of domain concepts, the paper proposes a concept clustering algorithm fusing the associations between concepts based on user-generated data. This method takes full advantage of user preferences derived from user-generated data, and effectively transforms the user preferences implied between concepts into the concept associations using mutual information. Considering that concept's own semantic is an important role for the similarity-based clustering, we propose a comprehensive semantic similarity measure method, which calculates the similarities in terms of distance, depth, density, coincidence degree and information content exploiting the principal analysis technique. On the basis of the above, we fuse the similarities of concepts with associations of that, making the original clustering results without concept associations get beneficial supplement in practical applications. The experimental results show the proposed method is effective and efficient. Compared with other existing methods, the precision is improved by 8%, the recall by 12%, and the F-Measure by 10%.