Fabiano Muniz Belém
Fabiano Muniz Belém has a total of 19 co-authors. Here are the top co-authors by number of shared publications:
- Alberto H. F. Laender
Universidade Federal de Minas Gerais, Belo Horizonte, Brazil
1 shared publication - Celso França
Federal University of Minas Gerais, Belo Horizonte, Brazil
1 shared publication - Claudio Lucio Do Val Lopes
Postgraduate Program in Mathematical and Computational Modeling, CEFET-MG, Belo Horizonte, Brazil
1 shared publication - Claudio Moisés Valiense de Andrade
Universidade Federal de Minas Gerais (UFMG)
1 shared publication - Eduardo Figueiredo
Federal University of Minas Gerais (UFMG)
1 shared publication - Fabricio Murai
Worcester Polytechnic Institute, Worcester, Massachusetts, USA
1 shared publication
Latest Publications
- Open Access
- 12 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 252-260
Engineering AI Agents for Clinical Workflows: A Case Study in Architecture, MLOps, and Governance
- 12 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 252-260
Open AccessEngineering AI Agents for Clinical Workflows: A Case Study in Architecture, MLOps, and Governance
- Open Access
- 31 October 2023
- Published by Sociedade Brasileira de Computação in Journal of Information and Data Management
- Vol. 14 (1)
Contextual Reinforcement, Entity Delimitation and Generative Data Augmentation for Entity Recognition and Relation Extraction in Official Documents
Transformer architectures have become the main component of various state-of-the-art methods for natural language processing tasks, such as Named Entity Recognition and Relation Extraction (NER+RE). As these architectures rely on semantic (contextual) aspects of word sequences, they may fail to accurately identify and delimit entity spans when there is little semantic context surrounding the named entities. This is the case of entities composed only by digits and punctuation, such as IDs and phone numbers, as well as long composed names. In this article, we propose new techniques for contextual reinforcement and entity delimitation based on pre- and post-processing techniques to provide a richer semantic context, improving SpERT, a state-of-the-art Span-based Entity and Relation Transformer. To provide further context to the training process of NER+RE, we propose a data augmentation technique based on Generative Pretrained Transformers (GPT). We evaluate our strategies using real data from public administration documents (official gazettes and biddings) and court lawsuits. Our results show that our pre- and post-processing strategies, when used co-jointly, allows significant improvements on NER+ER effectiveness, while we also show the benefits of using GPT for training data augmentation.
- 31 October 2023
- Published by Sociedade Brasileira de Computação in Journal of Information and Data Management
- Vol. 14 (1)
Open AccessContextual Reinforcement, Entity Delimitation and Generative Data Augmentation for Entity Recognition and Relation Extraction in Official Documents
Transformer architectures have become the main component of various state-of-the-art methods for natural language processing tasks, such as Named Entity Recognition and Relation Extraction (NER+RE). As these architectures rely on semantic (contextual) aspects of word sequences, they may fail to accurately identify and delimit entity spans when there is little semantic context surrounding the named entities. This is the case of entities composed only by digits and punctuation, such as IDs and phone numbers, as well as long composed names. In this article, we propose new techniques for contextual reinforcement and entity delimitation based on pre- and post-processing techniques to provide a richer semantic context, improving SpERT, a state-of-the-art Span-based Entity and Relation Transformer. To provide further context to the training process of NER+RE, we propose a data augmentation technique based on Generative Pretrained Transformers (GPT). We evaluate our strategies using real data from public administration documents (official gazettes and biddings) and court lawsuits. Our results show that our pre- and post-processing strategies, when used co-jointly, allows significant improvements on NER+ER effectiveness, while we also show the benefits of using GPT for training data augmentation.
- Top Cited PapersOpen Access
- 23 October 2023
- Published by Association for Computing Machinery (ACM)
- p. 159-167
Helping Fact-Checkers Identify Fake News Stories Shared through Images on WhatsApp
Digital news outlets have largely replaced traditional newspapers and television as the primary channels for information consumption in Brazil, and WhatsApp plays a crucial role in this context, by disseminating specific news stories. The creation of large public chat groups with numerous users and the ease of message forwarding have made the app popular among Brazilians as an affordable and immediate communication alternative. However, this has also put the platform in a central position in spreading misinformation campaigns. The platform’s closed architecture, protected by end-to-end encryption, poses a challenge for investigating and fact-checking WhatsApp content, hampering efforts to combat this problem. In this work, we explore automatic ranking-based strategies to propose a “fakeness score” model as a means to help fact-checking agencies identify fake news stories shared through images on WhatsApp. Based on the results, we design a tool and integrate it into a real system that has been used extensively for monitoring content during the 2018 Brazilian general election. Our experimental evaluation shows that this tool can reduce by up to 40% the amount of effort required to identify 80% of the fake news in the data when compared to current mechanisms practiced by the fact-checking agencies for the selection of news stories to be checked.
- 23 October 2023
- Published by Association for Computing Machinery (ACM)
- p. 159-167
Top Cited PapersOpen AccessHelping Fact-Checkers Identify Fake News Stories Shared through Images on WhatsApp
Digital news outlets have largely replaced traditional newspapers and television as the primary channels for information consumption in Brazil, and WhatsApp plays a crucial role in this context, by disseminating specific news stories. The creation of large public chat groups with numerous users and the ease of message forwarding have made the app popular among Brazilians as an affordable and immediate communication alternative. However, this has also put the platform in a central position in spreading misinformation campaigns. The platform’s closed architecture, protected by end-to-end encryption, poses a challenge for investigating and fact-checking WhatsApp content, hampering efforts to combat this problem. In this work, we explore automatic ranking-based strategies to propose a “fakeness score” model as a means to help fact-checking agencies identify fake news stories shared through images on WhatsApp. Based on the results, we design a tool and integrate it into a real system that has been used extensively for monitoring content during the 2018 Brazilian general election. Our experimental evaluation shows that this tool can reduce by up to 40% the amount of effort required to identify 80% of the fake news in the data when compared to current mechanisms practiced by the fact-checking agencies for the selection of news stories to be checked.
- 28 February 2023
- Published by Wiley in Software: Practice and Experience
- Vol. 53 (6) , 1438-1464
Dual analysis for helping developers to find collaborators based on co‐changed files: An empirical study
Summary: Software developers must collaborate at all stages of the software life‐cycle to create successful complex software systems. To enable this collaboration, social coding platforms, for example, GitHub, include an increasing number of tools to support collaboration. However, for large projects with hundreds of dynamic developers, such as several successful open–source projects, it can be complex to find developers with the same interest and familiarity and thus, gain suitable collaborations and new insights. In this context, resources and efforts may be wasted, discouraging many developers from contributing. Moreover, it can be costly to manage many contributions, which is another challenge for the maintainer who wants to take advantage of this small, timid, but valuable contribution made by a volunteer developer in a short time. In this context, this paper presents an empirical study aiming to evaluate two strategies to recommend collaborators based on co‐changed files. Inspired in the TF–IDF (Term Frequency–Inverse Document Frequency) weighting scheme established in the Information Retrieval field, these strategies first estimate the importance of relevant files modified by developers and use these estimates to represent each developer “profile”. As a second step, they estimate the similarity between developers using the Cosine metric, providing top‐ranked developers according to this measure as recommendations. We evaluated these strategies based on an extensive survey with 102 real–world developers. We observed that developers have interest and familiarity with the co‐changed files for all strategies evaluated. These considerations are of relevance because many opportunities for contributions to the project are linked to coding. Thus, theses results may indicate one less barrier for improving collaboration among developers. Overall, the strategies present an acceptance rate of up to 81%, contributing to the discovery of further collaborators.
- 28 February 2023
- Published by Wiley in Software: Practice and Experience
- Vol. 53 (6) , 1438-1464
Dual analysis for helping developers to find collaborators based on co‐changed files: An empirical study
Summary: Software developers must collaborate at all stages of the software life‐cycle to create successful complex software systems. To enable this collaboration, social coding platforms, for example, GitHub, include an increasing number of tools to support collaboration. However, for large projects with hundreds of dynamic developers, such as several successful open–source projects, it can be complex to find developers with the same interest and familiarity and thus, gain suitable collaborations and new insights. In this context, resources and efforts may be wasted, discouraging many developers from contributing. Moreover, it can be costly to manage many contributions, which is another challenge for the maintainer who wants to take advantage of this small, timid, but valuable contribution made by a volunteer developer in a short time. In this context, this paper presents an empirical study aiming to evaluate two strategies to recommend collaborators based on co‐changed files. Inspired in the TF–IDF (Term Frequency–Inverse Document Frequency) weighting scheme established in the Information Retrieval field, these strategies first estimate the importance of relevant files modified by developers and use these estimates to represent each developer “profile”. As a second step, they estimate the similarity between developers using the Cosine metric, providing top‐ranked developers according to this measure as recommendations. We evaluated these strategies based on an extensive survey with 102 real–world developers. We observed that developers have interest and familiarity with the co‐changed files for all strategies evaluated. These considerations are of relevance because many opportunities for contributions to the project are linked to coding. Thus, theses results may indicate one less barrier for improving collaboration among developers. Overall, the strategies present an acceptance rate of up to 81%, contributing to the discovery of further collaborators.