Studying Judicial Behaviour with Text Analysis

Abstract

Abstract This chapter provides an overview of computational text analysis techniques used to study judicial behaviour and decision-making. As legal texts become increasingly digitalized, scholars can draw on tools from machine learning and natural language processing to convert unstructured texts into quantitative data that is amenable to empirical analysis. A burgeoning field of computational legal analysis uses these tools to study the law and legal institutions. The chapter surveys common methods for representing legal documents. These methods vary in sophistication from simple word counts to more complex tools such as topic modelling and word embeddings. Each approach has strengths and weakness and makes trade-offs between richness of representation, interpretability, and dimensionality reduction. Extracting meaningful data from legal texts requires thoughtful choices about representation and preprocessing. The chapter discusses interpretation, bias, and domain-specific challenges as important considerations when applying text analysis to study courts. Overall, computational text analysis supplements traditional methods and opens new avenues for empirical legal scholarship.