Authors: Dor Bernsohn, Gil Semo, Yaron Vazana, Gila Hayat, Ben Hagag, Joel Niklaus, Rohit Saha, Kyryl Truskovskyi

Published on: February 06, 2024

Impact Score: 8.05

Arxiv code: Arxiv:2402.04335

Summary

• What is new: Fine-tuning models from the BERT family and conducting few-shot experiments with closed-source Large Language Models for legal NLP tasks.
• Why this is important: Detecting legal violations and associating them with affected individuals within unstructured textual data.
• What the research proposes: Constructed and used two datasets with Large Language Models, validated by expert annotators, specifically for class-action cases.
• Results: Achieved an F1-score of 62.69% in identifying legal violations and 81.02% in associating victims to these violations.

Technical Details

Technological frameworks used: BERT family models, open-source and closed-source LLMs

Models used: Fine-tuned and few-shot learning models

Data used: Two datasets constructed for legal NLP tasks, validated by domain experts

Potential Impact

Legal tech firms, law offices specializing in class-action lawsuits, legal documentation services

Want to implement this idea in a business?

We have generated a startup concept here: LegalTech Insight.