PeerJ Award Winners at CCS 2023

by | Jan 4, 2024 | Award Winner Interviews, Contributors

Davi Alves Oliveira University of Bahia State, Brazil. 

Can you tell us a bit about yourself and your research interests?

I am a doctoral candidate in the multidisciplinary and multi-institutional Postgraduate Program in Knowledge Diffusion (Programa de Pós-Graduação em Difusão do Conhecimento – PPGDC), affiliated with the Federal University of Bahia (UFBA), the University of Bahia State (UNEB), the Federal Institute of Education, Science and Technology of Bahia (IFBA), the State University of Feira de Santana (UEFS), the National Laboratory for Scientific Computing (LNCC), and the University Center Senai Cimatec. Additionally, I hold a permanent position as a professor at the University of Bahia State (UNEB).

Language has always been a subject of keen interest for me, leading me to explore various approaches, including Second Language Acquisition, Corpus Linguistics, Generative Linguistics, and Natural Language Processing. Currently, my focus lies in applying Network Science to the study of language, with a specific emphasis on analyzing the cohesion within written texts.

What first interested you in this field of research?

The ubiquity of language often leads us to take its complexity for granted. While pursuing my studies to become a teacher of English as a Foreign Language, I began exploring Linguistics and quickly realized that, despite its pervasive nature, language is not fully explained by our best models.

Later, as I delved into the field of Network Science, I discovered its diverse applications, particularly in modeling various language-related phenomena. The extensive possibilities and the novelty of the approach, specifically in computational network-based modeling of texts, were what initially captivated my interest.

Can you briefly explain the research you presented at CCS 2023?

At CCS2023, I presented part of an ongoing study exploring alternatives to representing sentences as subgraphs in semantic networks. Many studies traditionally represent sentences as cliques (i.e., complete subgraphs), overlooking their hierarchical structure. In our research, we propose two methods of edge pruning applied to isolated cliques representing sentences. These methods involve removing the weakest edges, based on metrics derived from word cooccurrence probabilities.

Preliminary results indicate that the networks constructed from these trimmed subgraphs retain certain characteristics of the networks constructed from cliques. Importantly, it offers the advantage of capturing some aspects of the hierarchical structure of sentences. These results are particularly valuable for the network-based analysis of text cohesion, focusing on the impact of sentences on the overall cohesion of texts.

How will you continue to build on this research?

The preliminary results presented at CCS2023 are part of a broader project aiming to model textual cohesion using networks. The forthcoming steps involve a comprehensive analysis to evaluate the efficacy of trimmed subgraphs in capturing the hierarchical structure of sentences. This will be done through comparisons with other methods. Additionally, the project aims to explore the utility of these networks of trimmed subgraphs in analyzing textual cohesion.

 

Hugo P. Maia Universidade Federal de Viçosa, Brazil. 

Can you tell us a bit about yourself and your research interests?

I have taken my bachelor’s and master’s degree in Physics at UFV, and though I considered studying other traditional areas of Physics, such as Relativity or Magnetism among others, Complex Systems drove my interest ever since a classmate explained to me the concept of Complex Networks. Since then I’ve been working at GISC (Grupo de Investigação de Sistemas Complexos) advised by Professors Silvio C. Ferreira and Marcelo L. Martins.

What first interested you in this field of research?

What I find most exciting in studying Complex Systems is its interdisciplinarity, often associating physics and math approaches to biology, ecology, epidemiology and social problems; the last one is the main interest of my recent research.

Can you briefly explain the research you presented at CCS 2023?

The research I presented at CCS2023 is an empirical study of polarization in online social networks. Recent years in Brazil, as in the rest of the world, have been marked by extreme polarization and radicalization. We have collected data from several tweets over recent years related to political debates in Brazil and analyzed them as interactions in a Multilayered Complex Network. It was noted that even though strong echo chamber effects, in which individuals talk preferentially to those who share similar opinions as yours, an anti-echo chamber effect, in which intensive interaction among individuals of opposing leaning is also observed. We conjecture that the latter plays an extreme relevance on how information propagates through a polarized community.

What are your next steps?

The next step is to analyze data from different time windows with different degrees of polarization. We expect to get more insights on how these anti-echo chamber interactions are formed and gauge their relevance. We also expect to validate spreading models to study how information flows to and from opposing sides of polarized communities.

Get PeerJ Article Alerts