Stance detection to probe the SCOTUS-public opinion problem.

This summer, I approached an age-old question in American democracy using natural language processing techniques.

The ideological leaning of the Supreme Court of the United States , as expressed through its decisions, seems to correlate with that of the public. In this project, we use computational linguistics techniques to investigate whether this association reflects strategic considerations or underlying political attitudes of the justices. To do this, we apply automated stance detection to the corpus of Supreme Court written decisions and oral arguments transcripts in an attempt to parse different justices’ views on both political issues and issues of broader jurisprudence. A practical byproduct of this work is a new, difficult stance detection dataset built on legal opinions, on which we show the efficacy of adapters as a means of tackling legal-specific tasks with general counterpart datasets.

YouTube Link to the video

Please sign in

If you are a registered user on Laidlaw Scholars Network, please sign in