Social Advertisability Analysis on Twitter

Twitter presents a nice opportunity for targeting advertisements that are contextually related to Twitter content. By virtue of the sparse and noisy text makes identifying the tweets for advertising a very hard problem. In this paper, we propose a novel and effective scheme to identify the tweets that can be targeted for advertisements. We firstly construct a multi-source corpus to collect more auxiliary information for advertisability analysis. We then build the LDA-based topic models to obtain the document-word distributions. We extract features according to these distributions and select contributing ones. Finally we train a logistic regression classifier to discriminate the advertisable tweets from unadvertisable ones. Extensive experiments on a representative real-word Twitter dataset demonstrate that our scheme can identify advertisable tweets effectively.