Adaptive Query Generation For Topic-Based Tweet Retrieval

Twitter has become a resource of great potential for analyzing opinion about hot topics. In this paper we show the methodology used for obtaning a corpus of Twitter messages related to the Spanish general elections of November 20, 2011. Given that access to Twitter messages is done through querying, we have studied various strategies for building such queries, trying to maximize the coverage. After experimenting with several approaches, we propose a graph-based method that allows retrieval of tweets related to a specific topic, dynamically adapting the queries to automatically include related topics that eventually arise. The obtained resource, very useful for, among others, sentiment analysis tasks, is publicy available for use.