Directed Social Regard: Surfacing Targeted Advocacy, Opposition, Aid, Harms, and Victimization in Online Media
Scott Friedman, Ruta Wheelock, Sonja Schmer-Galunder, Drisana Iverson, Jake Vasilakes + 4 more
TLDR
This paper introduces Directed Social Regard (DSR), a new NLP approach that identifies specific targets of mixed pro-social and anti-social sentiments in online text.
Key contributions
- Introduces Directed Social Regard (DSR) for multi-dimensional, multi-valence sentiment analysis.
- Uses two transformer models to detect span-level sentiment targets and score them on three social regard axes.
- Presents a data collection/annotation strategy and a transformer architecture for span-level scoring.
- Validates DSR model on third-party datasets, showing meaningful correlations with social science labels.
Why it matters
Existing sentiment analysis tools are limited in handling complex online language with mixed sentiments and multiple targets. DSR addresses this by providing a nuanced understanding of how pro-social and anti-social sentiments are directed. This advancement is crucial for analyzing influence operations, political rhetoric, and online harms, offering deeper insights into social dynamics.
Original Abstract
The language in online platforms, influence operations, and political rhetoric frequently directs a mix of pro-social sentiment (e.g., advocacy, helpfulness, compassion) and anti-social sentiment (e.g., threats, opposition, blame) at different topics, all in the same message. While many natural language processing (NLP) tools classify or score a text's overall sentiment as positive, neutral, or negative, these tools cannot report that positive and negative sentiments coexist, and they cannot report the target of those sentiments. This paper presents the Directed Social Regard (DSR) approach to multi-dimensional, multi-valence sentiment analysis, comprised of a pair of transformer-based models that (1) detects span-level targets of sentiment in a message and then (2) scores all spans within the message context along three (-1, 1) axes of regard that are motivated by social science theories of moral disengagement and moral framing. We present a data collection and annotation strategy for DSR dataset construction, a transformer-based architecture for span-level scoring, and a validation study with promising results. We apply the validated DSR model on six third-party datasets of online media and report meaningful correlations between DSR outputs and the labels and topics in these pre-existing social science datasets.
📬 Weekly AI Paper Digest
Get the top 10 AI/ML arXiv papers from the week — summarized, scored, and delivered to your inbox every Monday.