A large number of scientific publications are published continuously. It is difficult to distinguish essential publications from others. Bibliometric analysis allows such key publications to be identified typically months or years after publication, provided they have been cited sufficiently often. However, for current and high-frequency research areas, such as COVID-19, it is crucial to make this distinction shortly after publications are published. On social media, new publications are often shared and discussed within days of their release.
In the context of this thesis, an event-based system is presented that collects, processes, scores, and aggregates Twitter events that reference or discuss scientific publications.
Twitter events are automatically linked to referenced publications, bibliometric and social media data is further processed to generate a trend score to rank publications over time.
The developed proof-of-concept prototype stores data in a relational and a time-series database to provide trend data via a REST API. Moreover, the data is visualized with a modern web interface publicizing the current trends.