Banca de QUALIFICAÇÃO: ALEXANDRE GOMES DE LIMA

Uma banca de QUALIFICAÇÃO de DOUTORADO foi cadastrada pelo programa.
STUDENT : ALEXANDRE GOMES DE LIMA
DATE: 09/03/2023
TIME: 08:30
LOCAL: Google meet
TITLE:

Legal Rhetorical Role Labeling Supported by Transformer Models


KEY WORDS:

sentence encoding, BERT, judgment, sentence classification


PAGES: 91
BIG AREA: Ciências Exatas e da Terra
AREA: Ciência da Computação
SUBÁREA: Metodologia e Técnicas da Computação
SPECIALTY: Engenharia de Software
SUMMARY:

Legal AI, the application of Artificial Intelligence (AI) in the legal domain, is a research field that comprises several dimensions and tasks of interest. Like in other targeted application domains, one of the desired benefits is task automation, which increases the productivity of legal professionals and makes
law more accessible. Text is an important data source in the legal domain, therefore Legal AI has a great interest in Natural Language Processing (NLP) advances. This thesis concerns the automation of the Legal Rhetorical Role Labeling (RRL), a task that assigns semantic functions to sentences in legal documents. Legal RRl is a relevant task because it finds information that is useful in itself as well as for downstream tasks such as legal summarization and case law search. There are some factors that make Legal RRL a non-trivial task, even for humans: heterogeneity of documents’ source, lack of standards, domain expertise required, and subjectivity inherent to the task. These complicating factors and the large volume of legal documents justify the automation of the task. Such automation can be implemented as a sentence classification task: sentences are fed to a machine learning model that assigns a label, or class, for each sentence. Developing such models with a basis on Pre-trained Transformer Language Models (PTLM) is an obvious choice as PTLMs are the current state-of-the-art of many NLP tasks, including text classification. Although, in this thesis we highlight two main issues with works that exploit PTLMs to tackle the RRL task. The first one is the lack of works concerned with how to better deal with the idiosyncrasies of legal texts and with the usual small size and imbalance of Legal RRL datasets. Almost all related works simply employ the regular fine-tuning strategy to train models. The second issue is the poor harnessing of the intrinsic context exploitation  ability of PTLMs, which hampers models’ performance. This thesis proposal presents studies that aim to overcome such issues and hence advance the current state-of-the-art regarding the Legal RRL task. Out of these studies, two are complete, one is near completion, and two are in preparation. The results from these studies will be utilized to provide answers to the raised research questions.


COMMITTEE MEMBERS:
Presidente - 1671962 - EDUARDO HENRIQUE DA SILVA ARANHA
Interna - 1350250 - ANNE MAGALY DE PAULA CANUTO
Externo ao Programa - 2885532 - IVANOVITCH MEDEIROS DANTAS DA SILVA - UFRNExterno à Instituição - JOSÉ GUILLERMO MORENO
Externo à Instituição - TAOUFIQ DKAKI
Notícia cadastrada em: 07/03/2023 10:24
SIGAA | Superintendência de Tecnologia da Informação - (84) 3342 2210 | Copyright © 2006-2024 - UFRN - sigaa06-producao.info.ufrn.br.sigaa06-producao