In this paper, we describe our submission to the TREC 2011 Microblog track. We first use URLs as a clue to discover and remove the spam tweets. Then we use both Lucene and Indri to generate a ranked list of results for each query, together with their relevance scores. After that, we use the scores to find out useful hashtags relevant to the query, therefore some previously lower-ranked tweets can be discovered and are re-ranked higher. Query reformulation is considered in two of the four runs in our submissions.