Posts by Collection

JurisMap

An application that helps lawyers and judges find, compare and explore legal precedents with NLP, language models and visual analytics.

Zero-shot document classification for banking

A configurable framework that classifies documents with large language models, without labeled training data.

GenAI assistant for a private clinic

A GenAI digital assistant built as a client project for a private clinic.

ARTEH

An AWS-based platform for automated climate-risk assessment and TCFD-aligned reporting.

Prompt Engineering and Interaction Design course

A 14-hour course I designed at NOVA IMS on LLMs, prompting, context engineering, agentic coding tools and GenAI evaluation.

A Rapid Semi-automated Literature Review on Legal Precedents Retrieval

Published in EPIA Conference on Artificial Intelligence, 2022

This work uses text mining (TM), natural language processing (NLP), and data visualization methods to provide a semi-automated rapid literature review and identify how justice courts and legal practitioners may use AI to retrieve similar cases. Based on Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA), automation techniques were used to expedite the literature review. In this study, we confirmed the feasibility of automation tools for expediting literature reviews and provided an overview of the current research state on legal precedents retrieval.

Recommended citation: Silva, H., António, N., Bacao, F. (2022). A Rapid Semi-automated Literature Review on Legal Precedents Retrieval. In: Marreiros, G., Martins, B., Paiva, A., Ribeiro, B., Sardinha, A. (eds) Progress in Artificial Intelligence. EPIA 2022. Lecture Notes in Computer Science(), vol 13566. Springer, Cham. https://doi.org/10.1007/978-3-031-16474-3_5

Joining metadata and textual features to advise administrative courts decisions: a cascading classifier approach

Published in Artificial Intelligence and Law, 2023

Assessing the decisions made by regulatory bodies and courts is pivotal, given the vast influence they hold on the people. While machine learning (ML) may be used to predict such decisions, prevalent studies often overlook factors like consistency, real-world applicability, generality, and explainability. Our research introduces a unique two-stage cascade classifier model that harnesses both textual features and metadata from proceedings to improve performance. Utilizing the SHapley Additive exPlanations (SHAP) mechanism, our model remains transparent and explainable. With our approach, we’ve achieved a weighted F1 score of 0.900, outstripping baselines.

Recommended citation: Mentzingen, H., Antonio, N. & Lobo, V. Joining metadata and textual features to advise administrative courts decisions: a cascading classifier approach. Artif Intell Law (2023). https://link.springer.com/article/10.1007/s10506-023-09348-9

Automation of Legal Precedents Retrieval: Findings from a Literature Review

Published in International Journal of Intelligent Systems, 2023

This review article dives deep into the evolving landscape of automating the identification of legal precedents. It describes details of two “eras” of precedent retrieval automation, spotlighting the revolution powered by natural language processing (NLP) and machine learning (ML) until the use of Transformers. We also described the usual ML pipeline for precedent retrieval, the multiple techniques used, the data and geographies involved, and proposed a taxonomy for this field. A gap in validation and real-world deployments became evident, and a key question echoes: will courts of justice transform precedent searches through automation?

Recommended citation: Hugo Mentzingen, Nuno António, Fernando Bacao, "Automation of Legal Precedents Retrieval: Findings from a Literature Review", International Journal of Intelligent Systems, vol. 2023, Article ID 6660983, 22 pages, 2023. https://doi.org/10.1155/2023/6660983

Textual similarity for legal precedents discovery: Assessing the performance of machine learning techniques in an administrative court

Published in International Journal of Information Management Data Insights, 2024

We compare machine learning and language model techniques for measuring textual similarity between legal cases, using decisions from a Brazilian administrative court, to support the discovery of legal precedents.

Recommended citation: Hugo Mentzingen, Nuno António, Fernando Bacao, Marcio Cunha, "Textual similarity for legal precedents discovery: Assessing the performance of machine learning techniques in an administrative court", International Journal of Information Management Data Insights, vol. 4, no. 2, 100247, 2024. https://doi.org/10.1016/j.jjimei.2024.100247

Effectiveness in retrieving legal precedents: exploring text summarization and cutting-edge language models toward a cost-efficient approach

Published in Artificial Intelligence and Law, 2025

We study how text summarization interacts with language model embeddings when retrieving legal precedents, aiming for cost-efficiency. Using decisions from a Brazilian administrative court, we compare a legal-domain Transformer with OpenAI embeddings. Full texts with long-context embeddings retrieve best, while summaries built from part-of-speech tags offer a strong trade-off between performance and cost.

Recommended citation: Hugo Mentzingen, Nuno António, Fernando Bacao, "Effectiveness in retrieving legal precedents: exploring text summarization and cutting-edge language models toward a cost-efficient approach", Artificial Intelligence and Law, 2025. https://doi.org/10.1007/s10506-025-09440-2

Unveiling legal complexity: a systematic review on the visual analytics of legal corpora

Published in International Review of Law, Computers & Technology, 2026

A systematic review of how visual analytics has been applied to legal corpora, mapping the techniques, data and uses reported in the literature.

Recommended citation: Hugo Mentzingen, Nuno António, Fernando Bacao, "Unveiling legal complexity: a systematic review on the visual analytics of legal corpora", International Review of Law, Computers & Technology, vol. 40, no. 1, pp. 120-155, 2026. https://doi.org/10.1080/13600869.2025.2497630

Leveraging large language models for document classification in the banking sector

Published in International Journal of Data Science and Analytics, 2026

We propose a zero-shot document classification framework based on large language models, configurable by technical and non-technical users. Developed and validated on a real-world banking dataset, it classifies documents from user-provided label descriptions, without labeled training data.

Recommended citation: Rómulo Nogueira, Hugo Mentzingen, Nuno Garcia, "Leveraging large language models for document classification in the banking sector", International Journal of Data Science and Analytics, vol. 22, 2026. https://link.springer.com/article/10.1007/s41060-026-01077-x

Descriptive Methods of Data Mining

Undergraduate course, Nova IMS, 2021

This course addressed the topics and issues typically associated with “data mining” or “knowledge discovery”. Its primary goal is to allow students to gather skills for extracting information and knowledge from large databases. The skills developed include databases, data processing, prescriptive analysis, and data visualization.

Data Science for Marketing

Postgraduate course, Nova IMS, 2021

This course familiarizes students with Data science concepts, applications, and projects’ lifecycles. It involves statistics, data visualization, database systems, and machine learning. Students learn data preparation before building analytical models, such as data description, RFM, or association rules (e.g., market basket analysis).

Business Cases with Data Science

Masters course, Nova IMS, 2022

In this course, students were guided through six projects involving different topics in data science, such as language processing, recommendation systems, predictive analysis, and data visualization.

Machine Learning for Marketing

Course, NOVA IMS, 2024

Students apply the CRISP-DM process to marketing problems, from data preparation to regression, classification and clustering models, through three group projects.

Python Bootcamp Level II

Bootcamp, NOVA IMS, 2024

An intensive bootcamp that takes students from Python basics to the data manipulation skills used in machine learning courses.

Prompt Engineering and Interaction Design

Course, NOVA IMS, 2026

A 14-hour course I designed for students from product, business and management backgrounds. It covers how LLMs work, prompting and context engineering, agentic coding tools, and GenAI evaluation, and ends with a capstone in which students build and evaluate an AI-assisted application without writing code.