Real-Time Scientific Impact Prediction in Twitter.

Zhunchen Luo,Jun Chen,Xiao Liu

DOI: https://doi.org/10.1007/978-981-13-2922-7_7

2018-01-01

Abstract:As the number of scientific publication is getting larger and larger, scientific impact prediction has become an urgent need. However, traditional scientific impact prediction, which is mainly based on longtime accumulated citation networks, metadata and the whole text of papers, is relatively hysteretic and can hardly fit the rapid development of technology. Moreover, Twitter has become one of the most import channels to spread latest technique information because of its fast information spread speed. The advantage of publishing new messages in real-time can compensate the imperfections of traditional scientific impact prediction methods. Therefore, we propose a new approach to predict scientific impact in Twitter in real time before publishing the paper content. After filtering scholarly tweets (ST tweets), and extracting Tweet Scholar Blocks (TSBs) indicating metadata of papers to help predict scientific impact in real time, author social features, venue popularity features, and title features are exploited to predict whether the article will increase h-index of its first author after five years. Our model achieves an outstanding result that its best accuracy is 80.95%. The best feature conjunction consists of the sum of friends and followers of all the co-authors, followers count of the first author and title embeddings. And the amount of followers of all the co-authors is the most critical feature. Our finding reveals that Twitter has the potential to predict scientific impact in real time. We hope that real-time scientific impact prediction in Twitter can help researchers to expand their influences and more conveniently “stand on the shoulders of giants”.