Bonafide at LegalLens 2024 Shared Task: Using Lightweight DeBERTa Based Encoder For Legal Violation Detection and Resolution

TL;DR

This paper introduces lightweight DeBERTa-based systems for legal violation detection and resolution, achieving competitive performance in the LegalLens challenge and outperforming larger language models.

Contribution

The work presents novel lightweight DeBERTa-based models for legal violation detection and resolution, with improved accuracy over LLM baselines and public release of trained models.

Findings

NER system achieved 60.01% F1 score

NLI system achieved 84.73% F1 score

Ranked sixth and fifth on the LegalLens leaderboard

Abstract

In this work, we present two systems -- Named Entity Resolution (NER) and Natural Language Inference (NLI) -- for detecting legal violations within unstructured textual data and for associating these violations with potentially affected individuals, respectively. Both these systems are lightweight DeBERTa based encoders that outperform the LLM baselines. The proposed NER system achieved an F1 score of 60.01\% on Subtask A of the LegalLens challenge, which focuses on identifying violations. The proposed NLI system achieved an F1 score of 84.73\% on Subtask B of the LegalLens challenge, which focuses on resolving these violations by matching them with pre-existing legal complaints of class action cases. Our NER system ranked sixth and NLI system ranked fifth on the LegalLens leaderboard. We release the trained models and inference scripts.