For opinion finding task our method of the combination of 5 Windows method and Pseudo Relevance Feedback behaves well, achieving an improvement of over 20% on the baseline adhoc results. For the polarity task we develop two different methods. One is a classification method, and the other uses queries to retrieve positive and negative documents respectively. In Blog Distillation task, Pseudo Relevance Feedback method helps improve the result a little, however, since its dependence on the top 10 retrieval result, the method still need to be improved in order to get better result.