Paraphrase identification using weighted dependencies and word semantics