Large language models (LLMs) for legal analysis: RAG and beyond for optimizing domain adaptation in Portuguese legal domain

Barros, Tiago Mendonça Alencar

Publicação

Large language models (LLMs) for legal analysis: RAG and beyond for optimizing domain adaptation in Portuguese legal domain

2025-01-10Dissertação de mestrado

datacite.subject.fos	Ciências Sociais::Economia e Gestão
dc.contributor.advisor	Han, Qiwei
dc.contributor.author	Barros, Tiago Mendonça Alencar
dc.date.accessioned	2026-06-18T13:25:09Z
dc.date.available	2026-06-18T13:25:09Z
dc.date.issued	2025-01-10
dc.date.submitted	2024-12-17
dc.description.abstract	This study explores RAG systems tailored to the Portuguese legal domain, highlighting challenges in underrepresented languages. Fixed-size chunking strategies, particularly Token Text Splitter, 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. Self-reflection techniques show promising results, particularly for weaker LLMs. Techniques such as adding a pre-post translation proved to be an efficient technique for mitigating language bias.	eng
dc.identifier.tid	203927613
dc.identifier.uri	http://hdl.handle.net/10362/203859
dc.language.iso	eng
dc.relation	UID/ECO/00124/2013
dc.rights.uri	http://creativecommons.org/licenses/by/4.0/
dc.subject	Retrieval-augmented generation
dc.subject	RAG
dc.subject	Large language models
dc.subject	LLM
dc.subject	Artificial intelligence
dc.subject	AI
dc.subject	Hallucination
dc.subject	Question answering
dc.subject	RAG evaluation
dc.subject	Vector store
dc.subject	Chunking
dc.subject	Legal AI
dc.subject	Knowledge graph
dc.subject	GraphRAG
dc.subject	RDF
dc.subject	Graph-based reasoning
dc.subject	Self-assessment
dc.subject	Self-reflection
dc.subject	Multi-agent systems
dc.subject	MAS
dc.subject	Document reranking
dc.subject	Relevance ranking
dc.subject	Legal information retrieval
dc.subject	Portuguese legal retrieval
dc.subject	Machine translation
dc.subject	Natural language processing
dc.subject	LLM bias
dc.subject	Prompt engineering
dc.subject	Hierarchical indexing
dc.subject	Hierarchical retrieving
dc.subject	Chain-of-thought
dc.title	Large language models (LLMs) for legal analysis: RAG and beyond for optimizing domain adaptation in Portuguese legal domain	eng
dc.type	master thesis
dspace.entity.type	Publication
thesis.degree.name	A Work Project, presented as part of the requirements for the Award of a Master’s degree in Business Analytics from the Nova School of Business and Economics