Twitter sentiment analysis through transformers

NAISS 2023/22-22


NAISS Small Compute

Principal Investigator:

Nils Droste


Lunds universitet

Start Date:


End Date:


Primary Classification:

50601: Political Science (excluding Public Administration Studies and Globalisation Studies)



There is no clear scientific consensus about the effects of the herbicide glyphosate on human health and the environment. Yet, the societal debates about its regulation and economic consequences are heated. Social media (and more specifically Twitter) content constitutes a useful source of information and influence for policy makers, stakeholders and other actors when it comes to important socio-ecological issues such as the potentially harmful use of glyphosate in farming and gardening. Understanding the dynamics and (emotional) content of the discourse can raise awareness and help to take relevant actions. To map the discursive terrain over time, we will employ state of the art social media analysis algorithms that allow us to derive sentiment (like happiness and sadness), stances (like certainty and uncertainty), and latent topics from around 2.5 million tweets on glyphosate from 2006-2019 in all European languages. Thereby, the GlyphoSentiment project will contribute knowledge on how to tackle the socio-ecological challenges and improve our understanding of how science, policy, and social media discourse interact on a highly contested topic.