Sentiment Analysis in Altmetric

Sentiment analysis in Altmetric is a new feature that uses artificial intelligence (AI) and a Large Language Model (LLM) to understand the opinion expressed in online mentions of research outputs. Currently, this analysis is applied to selected social media posts only. The AI assigns a sentiment score to each post based on the language used, indicating whether the mention expresses a positive, negative, or neutral opinion about the research.

Sentiment data is displayed in Altmetric Explorer and on the Details Pages, helping users understand the context of online conversations about research.

The sentiment scores range from strongly negative (-3) to neutral (0) to strongly positive (3), with levels in between. Here’s a breakdown of the sentiment categories:

• -3 Strongly negative: Expresses strong criticism, warns against, or alerts about the mentioned paper.

• -2 Weak negative: Casts doubt, questions, cautions, or queries the research.

• -1 Uncertain negative: Includes satire, irony, humor, concern, or vague negative hints.

• 0 Neutral: Contains no sentiment towards the research, such as simply sharing a link or title.

• 1 Uncertain positive: For example, a human shares the title and link, possibly with a tag or hashtag.

• 2 Weak positive: Includes some commentary, suggests reading, or uses the research to support an argument without explicit praise.

• 3 Strong positive: Expresses strong recommendation, calls the research essential or a solution, or praises it as great work or evidence.

Sentiment analysis provides valuable context that goes beyond simple counts of mentions. Understanding the sentiment surrounding research can help you:

• Gain a deeper understanding of research reception: See how the public and professionals are responding to your research.

• Evaluate research impact: Determine if the engagement with your research is positive, indicating it's impactful and widely used, or negative, providing insights for future refinement.

• Inform future research directions: Understand what aspects of your research resonate or are questioned, providing valuable feedback for future work.

• Mitigate potential reputation risks: Quickly gauge public perception and address concerns if negative sentiment is detected.

• Demonstrate value to stakeholders and funders: Showcase not just the quantity but the quality of engagement with your research.

• Assess performance: Compare sentiment across different research outputs or benchmark against competitors.

• Develop targeted communication strategies: Understand which aspects of the research resonate most with different audiences.

• Identify key opinion leaders (KOLs): Find influential individuals expressing positive or negative sentiments.

Ultimately, sentiment analysis helps you move beyond anecdotal interpretations of mention counts and gain meaningful insights into the reception and impact of your research.

The sentiment breakdowns are often linked to the Mentions Tab. This allows you to easily click through to see the specific mentions that contributed to a particular sentiment category.

By utilizing these features, you can gain a richer understanding of the online conversation surrounding your research and make more informed decisions about your research and communication strategies.

Remember that sentiment analysis is currently available for selected social media data. We are continuously working to improve and expand this feature.

To learn more about the sentiment analysis in the Altmetric products, visit these pages:

Sentiment Analysis in the Explorer

Sentiment Analysis in Details Pages