Real-time personalized twitter search based on semantic expansion and quality model

The vast amount of information in social networks makes it difficult for users to find what they want, users may get drowned in the information flood. It is a challenging problem to retrieve the high quality and relevant information according to a user's searching query. Traditional methods for personalized search become insufficient in social networks due to the high velocity, topic variety, data sparseness and high sociability. To overcome those difficulties, we propose a novel framework for real-time personalized twitter search for twitter stream in this paper. Firstly, we develop a boolean logic keyword filter to enhance the accuracy. Then a tweet quality model is built to distinguish high quality tweets, it could improve the ranking performance. After that, we utilize an external search engine to implement query expansion, which could understand user preferences and interests properly. Our framework integrates the semantic features and social attributes which are utilized to make a comprehensive rank for a tweet. In addition, we adopt a dynamic strategy to push high quality and relevant tweets to a user automatically to avoid information overload. A thorough evaluation is conducted using real twitter stream data in TREC 2015, demonstrating a superior performance against competitive baselines in a variety of metrics.