An effective approach to tweets opinion retrieval

Opinion retrieval deals with finding relevant documents that express either a negative or positive opinion about some topic. Social Networks such as Twitter, where people routinely post opinions about almost any topic, are rich environments for opinions. However, spam and wildly varying documents makes opinion retrieval within Twitter challenging. Here we demonstrate how we can exploit social and structural textual information of Tweets and improve Twitter-based opinion retrieval. In particular, within a learning-to-rank technique, we explore the question of whether aspects of an author (such as the number of friends they have), information derived from the body of Tweets and opinionatedness ratings of Tweets can improve performance. Experimental results show that social features can improve retrieval performance. Retrieval using a novel unsupervised opinionatedness feature achieves comparable performance with a supervised method using manually tagged Tweets. Topic-related specific structured Tweet sets are shown to help with query-dependent opinion retrieval. Finally, we further verify the effectiveness of our approach for opinion retrieval in re-tagged TREC Tweets2011 corpus.