Combining statistics and semantics via ensemble model for document clustering