PhD Dissertation Defense: Michael Kranzlein

Title: “Unpacking Meaning with NLP: Legal Metalanguage Analysis and Long-Tail Calibration”

Metalanguage describes the reflexive capacity of language to be used simultaneously as a means of description and the target of description. In judicial opinions, metalanguage is frequent and of interest to legal scholars, yet metalanguage in judicial opinions, or more broadly, the legal domain, has not been characterized from a computational linguistics perspective.

This dissertation organizes my prior, published work as well as recent unpublished research around the construction of a corpus of legal metalanguage, the role of calibration for models and datasets with rare categories, and the application of transformer-based models to scale social science research efforts like legal content analysis of Supreme Court opinions.

Committee members:
Nathan Schneider (adviser)
Ophir Frieder
Lisa Singh
Dallas Card (University of Michigan)
Shomir Wilson (Penn State University)