ReLeVAnT: New Framework for Legal Text Classification
A study published on arXiv (2604.22292) presents ReLeVAnT, a framework designed for the binary classification of legal texts. This approach employs n-gram processing, contrastive score matching, and a simple neural network to categorize documents by their distinguishing characteristics, eliminating the need for structured metadata or significant computational resources. Notably, it necessitates only a single instance of keyword extraction for each corpus prior to the classification process.
Key facts
- arXiv paper 2604.22292
- ReLeVAnT framework for legal document binary classification
- Uses n-gram processing, contrastive score matching, shallow neural network
- One-time keyword extraction per corpus
- Avoids structured metadata and extensive computational power
Entities
Institutions
- arXiv