Adapting sentiment analysis for tweets linking to scientific papers

TL;DR

This paper investigates how sentiment analysis methods can be adapted to tweets linking to scientific papers, aiming to clarify Twitter's role in measuring research impact and improving altmetric evaluations.

Contribution

It introduces adapted sentiment analysis techniques tailored for academic tweets, providing insights into Twitter's function in scholarly communication and impact measurement.

Findings

Tweets mainly disseminate bibliographic info, not sentiments

Adapted methods improve sentiment detection in academic tweets

Results clarify Twitter's role in research impact assessment

Abstract

In the context of altmetrics, tweets have been discussed as potential indicators of immediate and broader societal impact of scientific documents. However, it is not yet clear to what extent Twitter captures actual research impact. A small case study (Thelwall et al., 2013b) suggests that tweets to journal articles neither comment on nor express any sentiments towards the publication, which suggests that tweets merely disseminate bibliographic information, often even automatically. This study analyses the sentiments of tweets for a large representative set of scientific papers by specifically adapting different methods to academic articles distributed on Twitter. Results will help to improve the understanding of Twitter's role in scholarly communication and the meaning of tweets as impact metrics.