Retain or Reframe? A Computational Framework for the Analysis of Framing in News Articles and Reader Comments

TL;DR

The paper tackles how framing in news articles interacts with reader comments at scale, addressing a gap where NLP analyses often treat source and response framing in isolation. It introduces a scalable computational framework that jointly analyzes frames in articles and comments, reconstructs dominant frames, and aligns responses with article topics. The FrAC corpus and a cross-domain frame classifier enable large-scale analysis across eleven topics and two outlets, revealing robust frame retention patterns and topic-specific reframing tendencies. The work demonstrates that audience reframing is common and systematically varies by frame type and topic, with implications for understanding media effects and for downstream NLP tasks such as persuasive text modeling and misinformation detection.

Abstract

When a news article describes immigration as an "economic burden" or a "humanitarian crisis," it selectively emphasizes certain aspects of the issue. Although \textit{framing} shapes how the public interprets such issues, audiences do not absorb frames passively but actively reorganize the presented information. While this relationship between source content and audience response is well-documented in the social sciences, NLP approaches often ignore it, detecting frames in articles and responses in isolation. We present the first computational framework for large-scale analysis of framing across source content (news articles) and audience responses (reader comments). Methodologically, we refine frame labels and develop a framework that reconstructs dominant frames in articles and comments from sentence-level predictions, and aligns articles with topically relevant comments. Applying our framework across eleven topics and two news outlets, we find that frame reuse in comments correlates highly across outlets, while topic-specific patterns vary. We release a frame classifier that performs well on both articles and comments, a dataset of article and comment sentences manually labeled for frames, and a large-scale dataset of articles and comments with predicted frame labels.

Figures (7)

• Figure 1: Overview of our framework. Bottom: Article with two comments illustrating reframing (Comment 1; changes the article frame) and frame retention (Comment 2; keeps the article frame).
• Figure 2: Predicted frame retention rates in reader comments based on regression-derived marginal effects. X- and Y-axes show predicted retention for SOCC and NYT, respectively. Retention varies by frame type, and the predictions correlate strongly across the two media outlets.
• Figure 3: Predicted topic retention rates in reader comments based on regression-derived marginal effects. Points above the diagonal line indicate higher predicted retention in NYT, while points below indicate higher predicted retention in SOCC. Topic names are shortened for readability.
• Figure 4: Reframing in the New York Times. (a) Complete reframing patterns where readers use frames that never appear anywhere in the original articles. (b) Selective reframing where readers emphasize secondary frames that differ from the dominant article frame but exist elsewhere in the article. Left percentages show dominant article frame distributions; right percentages show comment frame distributions among complete (a) / selective (b) reframing cases.
• Figure 5: An example from the annotation task on the deployed interface.