Multi-view Sentence Representation Learning