Learning from context: a mutual reinforcement model for Chinese microblog opinion retrieval