Mining network-level properties of Twitter altmetrics data

Abstract

Social networking sites play a significant role in altmetrics. While 90% of all altmetric mentions come from Twitter, the known microscopic and macroscopic properties of Twitter altmetrics data are limited. In this study, we present a large-scale analysis of Twitter altmetrics data using social network analysis techniques on the ‘mention’ network of Twitter users. Exploiting the network-level properties of over 1.4 million tweets, corresponding to 77,757 scholarly articles, this study focuses on the following aspects of Twitter altmetrics data: (a) the influence of organizational accounts; (b) the formation of disciplinary communities; (c) the cross-disciplinary interaction among Twitter users; (d) the network motifs of influential Twitter users; and (e) testing the small-world property. The results show that Twitter-based social media communities have unique characteristics, which may affect social media usage counts either directly or indirectly. Therefore, instead of treating altmetrics data as a black box, the underlying social media networks, which may either inflate or deflate social media usage counts, need further scrutiny.