Lucas Raniére Juvino Santos
Ph.D., Developer and Researcher
About
I am a computer scientist with research interests spanning social computing, natural language processing, and machine learning, including political polarization on social media, fact-checking, intelligent agents, and applied computer vision. I currently work on Judo-AI, applying computer vision to combat sports.
Academic Experience
Ph.D. in Computer Science
- Research topics: political polarization and radicalization in social media text, using large language models.
- Thesis: Large Language Models as Tools for Detection and Mitigation of Polarization on Social Media.
- December 2023 – August 2024: Visiting Ph.D. Researcher at Indiana University (Bloomington, Indiana).
Master's Degree in Computer Science
- Research topics: experimental software engineering and text-based traceability of software artifacts.
- Dissertation: Improving Traceability Recovery Between Bug Reports and Manual Test Cases.
Bachelor's Degree in Computer Science
- Research topics: text translation into sign language and digital inclusion of deaf people.
- Term paper: Use of Probabilistic Reasoning to Resolve Lexical Ambiguities in Portuguese.
Research Projects
Judo-AI — Applied Computer Vision for combat sports
- Computer vision
- Object detection
- Video segmentation
- Human action recognition
- Video understanding
Agents4Good — Intelligent Agents for Automatic Fact-checking
- Social computing
- AI agents
- LangChain and LangGraph
- Natural language processing
- Speech analysis
- Argument analysis
- Toxicity detection
Advanced Intelligent Channels — Applied ML for sales/support chat automation
- Natural language processing
- Machine learning
Publications
Automatic Classification of Judo Technique from Video: An Industry-Academia Collaboration
Presented at the 39th Conference on Graphics, Patterns and Images (SIBGRAPI '26), Goiânia, Brazil, September 29 – October 2, 2026. Paper not yet available in the proceedings.
Beyond Verdicts: Explainable Fact-Checking via a Linguistically-Grounded Multi-Agent Framework
Proceedings of the 37th ACM Conference on Hypertext and Social Media (HT '26). ACM, New York, NY, USA.
Who Is Who in Judo? Role-Aware Detection in Match Video
Presented at the 23rd Encontro Nacional de Inteligência Artificial e Computacional (ENIAC '26), Cuiabá, Brazil, October 19–22, 2026. Paper not yet available in the proceedings.
Can Large Language Models Effectively Mitigate Polarization in Social Media Text?
Proceedings of the 17th ACM Web Science Conference (WebSci '25). ACM, New York, NY, USA, 348–357.
Uniting Politics and Pandemic: a Social Network Analysis on the COVID Parliamentary Commission of Inquiry in Brazil
Proceedings of the Brazilian Symposium on Multimedia and the Web (WebMedia '22). ACM, New York, NY, USA, 99–107.
Improving Traceability Recovery Between Bug Reports and Manual Test Cases
Proceedings of the XXXIV Brazilian Symposium on Software Engineering (SBES '20). ACM, New York, NY, USA, 293–302.
Datasets and Models
Judo-AI 3 Classes Models (Weights from "best.pt")
Pre-trained YOLO object detection model weights for identifying three classes in judo matches: referee, blue athlete, and white athlete.
[Tweets] 2022 Brazilian Presidential Elections
A dataset of over 7 million tweets from 2022 collected to study political discourse during Brazil's presidential election.
[Tweets] 2023 Brazilian Early Political Events
A dataset of over 13.9 million Portuguese-language tweets from early 2023 for NLP and social network research on political discourse.
[Dataset] Tweets about COVID-19 Brazilian PCI
A dataset of roughly 3.4 million tweets collected over 26 weeks regarding Brazil's COVID-19 Parliamentary Commission of Inquiry.