Banca de QUALIFICAÇÃO: GISLIANY LILLIAN ALVES DE OLIVEIRA

Uma banca de QUALIFICAÇÃO de DOUTORADO foi cadastrada pelo programa.
STUDENT : GISLIANY LILLIAN ALVES DE OLIVEIRA
DATE: 10/08/2023
TIME: 14:00
LOCAL: Remoto
TITLE:

Textual Analysis of Legislative Documents and Parliamentary Speeches using Knowledge Graphs and Graph Neural Networks

KEY WORDS:

Legislative texts; Natural Language Processing; Knowledge Graphs; Graph Neural Networks.

PAGES: 47
BIG AREA: Engenharias
AREA: Engenharia Elétrica
SUMMARY:

The Legislative Branch has, as one of its most prominent tasks, the creation of laws that guide society. To fulfill this function, public institutions adopt the legislative process, a set of procedures for the creation of legal norms, which includes stages where the texts of propositions need to be examined, debated, revised, and eventually deliberated upon. These tasks are usually costly, as creating a new proposal requires considering its relationship with existing legislation and other proposals being processed in the legislative body. Additionally, a more technical writing style is inherent to texts in this field, which also tend to be longer. Therefore, the same broad scope and volume of data that make certain legislative process procedures onerous in terms of human effort pose a tangible challenge for Artificial Intelligence (AI), particularly in the field of Natural Language Processing (NLP). With widely disseminated methods and models in the literature, NLP has also benefited from the increasing adoption of Knowledge Graphs, which can inject structured information from their most meaningful representation of the data. Furthermore, the recent development of Graph Neural Networks (GNNs) has increased their capabilities and expressive power. In this scenario, this work proposes an approach to transpose legislative texts into the domain of Knowledge Graphs and subsequently use Graph Neural Networks to perform the textual analyses inherent to the legislative process. Given that deep learning tasks tend to be computationally expensive, the design of a pruning technique for Knowledge Graphs is also part of the scope of this project. The data used consists of the texts of legislative proposals and the speeches from parliamentary meetings, both provided by the Legislative Assembly of Rio Grande do Norte (ALRN). To assess the feasibility of the proposed approach, a series of quantitative experiments will be conducted using data from the legislative branch of Rio Grande do Norte, involving exploratory analyses and the training and validation of Graph Neural Networks models. Finally, the results of the models will be compared with existing approaches in the literature, and the effectiveness of graph pruning technique in reducing computational cost without compromising model performance will be evaluated.

COMMITTEE MEMBERS:
Presidente - 2885532 - IVANOVITCH MEDEIROS DANTAS DA SILVA
Interno - 1153006 - LUIZ AFFONSO HENDERSON GUEDES DE OLIVEIRA
Externo ao Programa - 2249146 - CARLOS MANUEL DIAS VIEGAS - UFRN