Sentiment and objectivity in Iranian state-sponsored propaganda on Twitter

Michael Barrows, Ella Haig, Dara Conduit

Research output: Contribution to journalArticlepeer-review

38 Downloads (Pure)


In 2016, Russia attempted to use social media to influence the outcome of the U.S. presidential election, highlighting the potential real-world impacts of state-led online misinformation campaigns. Misinformation on social media is a growing concern, especially in the areas of politics and medicine, given their impact not only at the individual level but also for society as a whole. In this article, we investigate the potential to automatically label and detect the polarity (positive, neutral, or negative) of Iranian state-sponsored propaganda tweets on the Iranian nuclear deal. The SentiWordNet lexicon is used to automatically assign a polarity label and an objectivity score to each tweet. Using the labels, five machine learning algorithms are used to create polarity detection models. The experimental results show that the best performing models correctly identify polarity in approximately 77% of the tweets.
Original languageEnglish
Number of pages10
JournalIEEE Transactions on Computational Social Systems
Early online date22 May 2023
Publication statusEarly online - 22 May 2023


  • Social networking (online)
  • Blogs
  • Sentiment analysis
  • Fake news
  • Voting
  • Filtering
  • Machine learning algorithms

Cite this