Author(s): Wollny, Carolyn Svea
Date: 2025
Persistent ID: http://hdl.handle.net/10362/186944
Origin: Repositório Institucional da UNL
Subject(s): Retrieval-Augmented Generation; RAG; Large Language Models; LLM; Artificial Intelligence; AI; Hallucination; Question answering; RAG evaluation; Vector store; Chunking; Legal AI; Document reranking; Relevance ranking; Legal information retrieval; Portuguese legal retrieval; Domínio/Área Científica::Ciências Sociais::Economia e Gestão
This study explores RAG systems tailored to the Portuguese legal domain, highlighting challenges in underrepresented languages. Fixed-size chunking strategies, particularly TokenTextSplitter, were found to be most effective, while more advanced techniques like Recursive and Semantic splitting showed little benefits. Larger chunk sizes improved retrieval accuracy and answer quality, though the impact of chunk overlap remains inconclusive. Although reranking techniques have been shown to improve retrieval in previous research this may only be true for large and diverse datasets.
