Exploratory Search on Twitter Utilizing User Feedback and Multi-Perspective Microblog Analysis

In recent years, besides typical information retrieval, a broader concept of information exploration – exploratory search - is emerging into the foreground. In addition, more and more valuable information is presented in microblogs on social networks. We propose a new method for supporting the exploratory search on the Twitter social network. The method copes with several challenges, namely brevity of microblogs called tweets, limited number of available ratings and the need to process the recommendations online. In order to tackle the first challenge, the representation of microblogs is enriched by information from referenced links, topic summarization and affect analysis. The small number of available ratings is raised by interpreting implicit feedback trained by feedback model during browsing. Recommendations are made by a preference model that models user’s preferences over tweets. The evaluation shows promising results even when navigating in the space of brief pieces of information, making recommendations based only on a small number of ratings, and by optimizing the models to process in real time.