24 April 2023
09:00 Master's Defense IC3 Auditorium
Theme
Revealing gender biases in court decisions with natural language processing
Student
Raysa Masson Benatti
Advisor / Teacher
Esther Luna Colombini - Co-advisor: Sandra Eliza Fontes de Avila
Brief summary
Data from the social sciences are commonly produced in the form of digital text, which motivates its use as a source for natural language processing methods. Researchers and professionals have been developing and using artificial intelligence techniques to collect, process and analyze documents in the legal field, especially in tasks such as summarizing and classifying texts. In this scenario, we identify an underexplored potential of using natural language processing to deal with human rights issues, in the context of artificial intelligence aimed at social good. In particular, we turn to identifying institutional gender biases of Brazilian courts, especially in cases related to domestic violence. Qualitative and quantitative methods from the social sciences have been used to study this question; however, semantic-based techniques for automatic text analysis could reveal such biases in court decisions on a larger scale. Therefore, we propose a protocol based on natural language processing to collect and analyze judicial decisions issued by the Court of Justice of São Paulo; such protocol relies on supervised classification tasks with networks based on attention mechanisms. Our approach can help specialists in areas such as Law, Gender Studies and Public Policy to explore new possibilities for analysis in their domain –- as well as providing insights into the use of natural language processing techniques and tools.
Examination Board
Headlines:
| Esther Luna Colombini | IC / UNICAMP |
| Luanna Tomaz de Souza | ICJ/UFPA |
| Rodrigo Frassetto Nogueira | NeuralMind |
Substitutes:
| Nádia Félix Felipe da Silva | INF/UFG |
| Roberto de Alencar Lotufo | FEEC / UNICAMP |