Banca de QUALIFICAÇÃO: PATRICIA DIAS DOS SANTOS

Uma banca de QUALIFICAÇÃO de DOUTORADO foi cadastrada pelo programa.
STUDENT : PATRICIA DIAS DOS SANTOS
DATE: 24/11/2022
TIME: 10:00
LOCAL: por participação remota em https://conferenciaweb.rnp.br/webconf/denise-11
TITLE:

Stance Detection and Twitter Users Automatic Labeling on Polarized and Politically Controversial Issues

PAGES: 110
BIG AREA: Ciências Exatas e da Terra
AREA: Ciência da Computação
SUBÁREA: Metodologia e Técnicas da Computação
SUMMARY:

Stance detection is a research problem related to social network analysis, natural language processing, machine learning, and information retrieval. A text is examined to determine a person's stance regarding a topic or subject that is explicitly or implicitly mentioned. Labels such as Favorable, Contrary, and Neutral are commonly used when rating positioning. 

People often express their outlook on social and political issues via social media. However, making predictions about their stance through manual labeling can be difficult. As the volume of social media posts is too large to be analyzed manually, computational methods are needed to identify positioning. 

This thesis proposes three main contributions: the development of a method for detecting stances and automatic labeling of Twitter users, the creation of a labeled corpus for the stance detection using retweets on politically controversial topics, and finally, the evaluation of the quality of the method, as well as the distribution of data in the generated corpus. 

Only a few corpora in Portuguese have positioning labels annotated to their data. As annotated data are not readily available, we created our own dataset from 13.4 million retweets collected over 22 weeks about the Covid-19 CPI. From the application of the model, it was possible to label over 740 thousand users, with minimal human intervention. Statistical analysis of the generated corpus showed that the final distribution of labels between opposing and favorable retweets is relatively similar. However, the opposite users, despite being in smaller numbers, do on average much more retweets than their counterparts. Another interesting observation is that the ratio of users/retweets against and in favor varied each week.

BANKING MEMBERS:
Presidente - Interno ao Programa - 2976815 - DENISE HIDEKO GOYA
Membro Titular - Examinador(a) Externo ao Programa - 1545036 - CLAUDIO LUIS DE CAMARGO PENTEADO
Membro Titular - Examinador(a) Externo ao Programa - 1849928 - CARLOS DA SILVA DOS SANTOS
Membro Suplente - Examinador(a) Externo à Instituição - KARIN BECKER - UFRGS
Membro Suplente - Examinador(a) Externo à Instituição - JULIO CESAR SOARES DOS REIS - UFV