top of page

Frame-sentiment dynamics and evolution in U.S. climate news:
Semi-supervised machine learning and panel data analysis

  • U.S. climate coverage from NYT and WSJ over 27 years, from 1996 to 2023 (56,475 articles).

  • Guided LDA topic modeling, Transformer-based sentiment analysis, and panel data analysis.

  • The results highlight the predominance of market-driven solutions in the media versus declining narratives seeking climate justice and global equity.

  • The methodological contribution combines deductive and inductive approaches to identify news frames to inform future public perception and policy-making strategies.

Guided LDA Topic Modeling (1)

Inductive pattern recognition.

Determine 40 keywords as seed words per frame,

considering visual distance.

​

3D 18topics_lda_umap_visualization.png

Guided LDA Topic Modeling (2)

Identifying Four Frames

Frames_lda_pca_visualization copy.png

​​

DATA

Proportion of Frames Over Time_wide.png
Sentiment Scores per Source_wide.png

​Selected Finding

Summary

©2017 BY BYUNG WOOK (BK) KIM. PROUDLY CREATED WITH WIX.COM

bottom of page