The Legal NLP Unit at Georgia Tech, led by Dr. Charlotte Alexander, gives students the opportunity to apply natural language processing to pressing issues in law and society. Sentencias extracts and classifies procedural events from Spanish-language court rulings in the Dominican Republic, helping to diagnose judicial delays. DNO uses document classification and transformer models to analyze U.S. Directors and Officers (D&O) insurance litigation, identifying key legal issues such as late notice and claim disputes. The NLP summarization project is focused on collaborating with the civil rights litigation clearinghouse to apply state of the art LLMs to summarize court documents to help aid law students in their work. These projects offer hands-on experience at the intersection of machine learning, legal reasoning, and social impact.
Projects
Directors and Officers Insurance (DNO)
NLP Project
This research project uses Natural Language Processing (NLP) to study Directors and Officers (D&O) insurance disputes from federal and state courts. The team is building machine learning models that can automatically read legal documents. These models extract the arguments that insurance companies and claimants make in coverage disputes. The system processes thousands of court filings to identify common patterns in coverage issues like late notification problems and disagreements over claim definitions. The goal is to create an automated system that reveals how D&O insurance actually works in practice. This will provide useful insights for researchers and insurance professionals. It will also reduce the time-consuming work of manually reviewing legal documents.
To learn more, check out:
https://humanaugmentedanalyticsgroup.miraheze.org/wiki/NLPSentencias
NLP Project
This research project is developing an automated system to analyze Spanish-language court decisions from the Dominican Republic, transforming lengthy unstructured legal documents into organized timelines of procedural events. The team created SLLED, a machine learning pipeline that first uses specialized Named Entity Recognition models to identify and extract dates from court rulings, then employs large language models to classify what type of legal event occurred on each date, e.g.: scheduled hearings, postponed hearings, or filing deadlines. The system includes an ontology verification component that checks whether the identified events make logical sense together and prompts the model to reconsider if inconsistencies are found. By processing labor court cases and creating structured event logs, the researchers aim to help Dominican Republic courts identify where procedural delays occur most frequently.
To learn more, check out:
https://humanaugmentedanalyticsgroup.miraheze.org/wiki/NLP