ReLeVAnT: Relevance Lexical Vectors for Accurate Legal Text Classification

2026-04-24

AI Summary

This work addresses the challenge of efficient binary classification of unstructured legal documents by proposing a lightweight approach that eschews reliance on metadata or large language models. The method leverages n-gram feature extraction and contrastive score matching to identify discriminative keywords, which are then fed into a shallow neural network to construct a low-computational-cost classification framework. Evaluated on the LexGLUE benchmark, the proposed model achieves 99.3% accuracy and 98.7% F1 score, substantially enhancing both the efficiency and precision of legal text relevance determination. This advancement offers a practical and effective solution for deploying legal AI applications in resource-constrained environments.

Technology Category

• Natural Language Processing: Text Classification & Sentiment Analysis
• Machine Learning: Multi-class/Multi-label Learning & Extreme Classification
• Computer Vision: Language and Vision

Application Category

• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Web Mining and Content Analysis: Normalization, clustering, classification, and summarization of Web text
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

The classification of legal documents from an unstructured data corpus has several crucial applications in downstream tasks. Documents relevant to court filings are key in use cases such as drafting motions, memos, and outlines, as well as in tasks like docket summarisation, retrieval systems, and training data curation. Current methods classify based on provided metadata, LLM-extracted metadata, or multimodal methods. These methods depend on structured data, metadata, and extensive computational power. This task is approached from a perspective of leveraging discriminative features in the documents between classes. The authors propose ReLeVAnT, a framework for legal document binary classification. ReLeVAnT utilises n-gram processing, contrastive score matching, and a shallow neural network as the primary drivers for discriminative classification. It leverages one-time keyword extraction per corpus, followed by a shallow classifier to swiftly and reliably classify documents with 99.3% accuracy and 98.7% F1 score on the LexGLUE dataset.

Problem

• legal text classification
• unstructured data
• binary classification
• document relevance
• metadata-free

Innovation

• n-gram processing
• contrastive score matching
• shallow neural network
• legal text classification
• discriminative feature extraction