Anderson Almeida Ferreira
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Anderson Almeida Ferreira has a total of 26 co-authors. Here are the top co-authors by number of shared publications:
- Ítalo M. Pereira
Federal University of Ouro Preto; Federal Institute of Minas Gerais
3 shared publications - Alberto H. F. Laender
Universidade Federal de Minas Gerais, Belo Horizonte, Brazil
2 shared publications Adriana C. F. Viana
Psychology Department, Universidade Federal de São João del-Rei, São João del-Rei, Brazil
1 shared publicationAlan Filipe Santana
Departamento de Ciência da Computação, Universidade Federal de Minas Gerais, 31270-901 Belo Horizonte, Brazil
1 shared publicationAna Cláudia Machado
Computer Science Department, Universidade Federal de São João del-Rei, São João del-Rei, Brazil
1 shared publicationAntônio Pereira de Souza Júnior
Ana Health, São Paulo, Brazil
1 shared publication
Latest Publications
- Open Access
- 27 August 2025
- Published by Springer Nature in Scientific Reports
- Vol. 15 (1) , 1-15
A comprehensive qualitative analysis of patient dialogue summarization using large language models applied to noisy, informal, non-English real-world data
This study evaluates the ability of Large Language Models (LLMs) to summarize real-world dialogues between patients and the healthcare team of an e-health company that provides digital healthcare services, primarily communicating via WhatsApp. The team needs quick access to patient information to deliver accurate and personalized responses. Summarizing past messages is the approach examined here, aiming for concise, non-redundant, and truthful summaries that capture the main dialogue characteristics despite facing (real-world) noisy and informal content in an under-represented language - Portuguese. To do so, we collected an anonymized Portuguese dataset of WhatsApp messages exchanged between patients and the healthcare team. Dialogue quality was assessed for size, readability, and correctness before generating summaries with LLaMA3 and Qwen2 using specific prompts. Volunteers evaluated these summaries on coverage, relevance, redundancy, and veracity using a 5-point Likert scale. Our qualitative and quantitative experimental results indicate that LLMs can produce effective summaries of dialogues between patients and healthcare teams, even when faced with low-quality data in an underrepresented language. This is a surprising result due to the challenging scenario. Among the tested LLMs, LLaMA3 demonstrated a slight edge over QWen2 in coverage and veracity among the evaluated methods. Our results demonstrate a potential to build real-world practical services to assist healthcare professionals in responding to patient messages with agility, clarity, and cohesion, enhancing both communication efficiency and patient satisfaction. Ultimately, the advocated approach could significantly improve the landscape of online healthcare communication, particularly in resource-constrained settings like Brazil, where access to primary care is limited.
- 27 August 2025
- Published by Springer Nature in Scientific Reports
- Vol. 15 (1) , 1-15
Open AccessA comprehensive qualitative analysis of patient dialogue summarization using large language models applied to noisy, informal, non-English real-world data
This study evaluates the ability of Large Language Models (LLMs) to summarize real-world dialogues between patients and the healthcare team of an e-health company that provides digital healthcare services, primarily communicating via WhatsApp. The team needs quick access to patient information to deliver accurate and personalized responses. Summarizing past messages is the approach examined here, aiming for concise, non-redundant, and truthful summaries that capture the main dialogue characteristics despite facing (real-world) noisy and informal content in an under-represented language - Portuguese. To do so, we collected an anonymized Portuguese dataset of WhatsApp messages exchanged between patients and the healthcare team. Dialogue quality was assessed for size, readability, and correctness before generating summaries with LLaMA3 and Qwen2 using specific prompts. Volunteers evaluated these summaries on coverage, relevance, redundancy, and veracity using a 5-point Likert scale. Our qualitative and quantitative experimental results indicate that LLMs can produce effective summaries of dialogues between patients and healthcare teams, even when faced with low-quality data in an underrepresented language. This is a surprising result due to the challenging scenario. Among the tested LLMs, LLaMA3 demonstrated a slight edge over QWen2 in coverage and veracity among the evaluated methods. Our results demonstrate a potential to build real-world practical services to assist healthcare professionals in responding to patient messages with agility, clarity, and cohesion, enhancing both communication efficiency and patient satisfaction. Ultimately, the advocated approach could significantly improve the landscape of online healthcare communication, particularly in resource-constrained settings like Brazil, where access to primary care is limited.
- Open Access
- 17 January 2025
- Published by Sociedade Brasileira de Computação in Journal of the Brazilian Computer Society
- Vol. 31 (1) , 598-612
A sentence similarity-based approach for enhancing entity linking
Entity linking involves associating mentions of entities in natural language texts, such as references to people or locations, with specific entity representations in knowledge graphs like DBpedia or Wikidata. This process is essential in natural language processing tasks, as it aids in disambiguating entities in unstructured data, thereby improving comprehension and semantic processing. However, entity linking faces challenges due to the complexity and ambiguity of natural languages, as well as discrepancies between the forms of textual entity mentions and entity representations. Building upon our previous work, this study extends the E-BELA --Enhanced Embedding-Based Entity Linking Approach, which is based on literal embeddings. We extend our previous work by evaluating E-BELA using a new dataset, conducting a comprehensive analysis of failure cases and limitations, and providing further discussion of our results. E-BELA associates mentions and entity representations using a similarity or distance metric between vector representations of them in a shared vector space. The results suggest that our approach achieves comparable performance to other state-of-the-art methods, while employing a much simpler model, contributing to the field of natural language processing.
- 17 January 2025
- Published by Sociedade Brasileira de Computação in Journal of the Brazilian Computer Society
- Vol. 31 (1) , 598-612
Open AccessA sentence similarity-based approach for enhancing entity linking
Entity linking involves associating mentions of entities in natural language texts, such as references to people or locations, with specific entity representations in knowledge graphs like DBpedia or Wikidata. This process is essential in natural language processing tasks, as it aids in disambiguating entities in unstructured data, thereby improving comprehension and semantic processing. However, entity linking faces challenges due to the complexity and ambiguity of natural languages, as well as discrepancies between the forms of textual entity mentions and entity representations. Building upon our previous work, this study extends the E-BELA --Enhanced Embedding-Based Entity Linking Approach, which is based on literal embeddings. We extend our previous work by evaluating E-BELA using a new dataset, conducting a comprehensive analysis of failure cases and limitations, and providing further discussion of our results. E-BELA associates mentions and entity representations using a similarity or distance metric between vector representations of them in a shared vector space. The results suggest that our approach achieves comparable performance to other state-of-the-art methods, while employing a much simpler model, contributing to the field of natural language processing.
- Open Access
- 17 November 2024
- Published by Sociedade Brasileira de Computação
- p. 153-160
An Ultra-Fine Entity Typing Method Based on Ensemble of Large Language Models
Ultra-fine Entity Typing (UFET) is an evolution of Named Entity Recognition (NER) that proposes the classification of entities at an ultra-fine level, based on a set of free-form phrases that adequately and comprehensively describe them. Increasing the number of labels while maintaining performance remains a challenge, and most current models are capable of classifying entities into a limited set of types. The proven efficiency of Large Language Models (LLMs) in improving performance in Natural Language Processing tasks presents an opportunity to enhance UFET results. In this sense, this paper proposes an ensemble method of fine-tuned public LLMs aiming to maximize the efficiency of entity classification at the ultra-fine level and expand the set of labels. Experiments show the effectiveness of the proposed method, which outperformed baselines metrics in almost all scenarios and improved the set of types.
- 17 November 2024
- Published by Sociedade Brasileira de Computação
- p. 153-160
Open AccessAn Ultra-Fine Entity Typing Method Based on Ensemble of Large Language Models
Ultra-fine Entity Typing (UFET) is an evolution of Named Entity Recognition (NER) that proposes the classification of entities at an ultra-fine level, based on a set of free-form phrases that adequately and comprehensively describe them. Increasing the number of labels while maintaining performance remains a challenge, and most current models are capable of classifying entities into a limited set of types. The proven efficiency of Large Language Models (LLMs) in improving performance in Natural Language Processing tasks presents an opportunity to enhance UFET results. In this sense, this paper proposes an ensemble method of fine-tuned public LLMs aiming to maximize the efficiency of entity classification at the ultra-fine level and expand the set of labels. Experiments show the effectiveness of the proposed method, which outperformed baselines metrics in almost all scenarios and improved the set of types.
- Open Access
- 14 October 2024
- Published by Sociedade Brasileira de Computação
- p. 115-123
E-BELA: Enhanced Embedding-Based Entity Linking Approach
Entity linking is the process of connecting mentions of entities in natural language texts, such as references to people or places, to specific entities in knowledge graphs, such as DBpedia or Wikidata. This process is crucial in the natural language processing tasks since it facilitates disambiguating entities in unstructured data, enhancing understanding and semantic processing. However, entity linking faces challenges due to the complexity and ambiguity of natural languages, as well as the discrepancy between the form of textual entity mentions and entity representations. Considering that entity mentions are in natural language and entity representations in knowledge graphs have object nodes that describe them in the same way, in this work we propose E-BELA, an effective approach based on literal embeddings. We aim to put close vector representations of mentions and entities in a vector space, allowing linking of mentions and entities by using a similarity or distance metric. The results demonstrate that our approach outperforms previous ones, contributing to the field of natural language processing.
- 14 October 2024
- Published by Sociedade Brasileira de Computação
- p. 115-123
Open AccessE-BELA: Enhanced Embedding-Based Entity Linking Approach
Entity linking is the process of connecting mentions of entities in natural language texts, such as references to people or places, to specific entities in knowledge graphs, such as DBpedia or Wikidata. This process is crucial in the natural language processing tasks since it facilitates disambiguating entities in unstructured data, enhancing understanding and semantic processing. However, entity linking faces challenges due to the complexity and ambiguity of natural languages, as well as the discrepancy between the form of textual entity mentions and entity representations. Considering that entity mentions are in natural language and entity representations in knowledge graphs have object nodes that describe them in the same way, in this work we propose E-BELA, an effective approach based on literal embeddings. We aim to put close vector representations of mentions and entities in a vector space, allowing linking of mentions and entities by using a similarity or distance metric. The results demonstrate that our approach outperforms previous ones, contributing to the field of natural language processing.
- Open Access
- 25 September 2023
- Published by Sociedade Brasileira de Computação
- p. 179-182
Automatic Disambiguation of Author Names: Foundations, Methods and Open Issues
This tutorial is based on our book “Automatic Disambiguation of Author Names in Bibliographic Repositories” and aims to spread the problem and its challenges among the SBBD community. Author name ambiguity problem occurs when an author publishes works under distinct names or distinct authors publish works under similar names. This problem may be caused by a number of reasons, including the lack of standards and common practices, and the decentralized generation of bibliographic content. In this tutorial, we intend to present an ample view on the automatic disambiguation of author names. We start by discussing its motivational issues, defining the author name disambiguation task and presenting its foundations. Next, we describe some methods proposed by our research group, as well as some recent approaches to author name disambiguation. Finally, we discuss open issues. This tutorial is based on our book “Automatic Disambiguation of Author Names in Bibliographic Repositories” and aims to spread the problem and its challenges among the SBBD community. Author name ambiguity problem occurs when an author publishes works under distinct names or distinct authors publish works under similar names. This problem may be caused by a number of reasons, including the lack of standards and common practices, and the decentralized generation of bibliographic content. In this tutorial, we intend to present an ample view on the automatic disambiguation of author names. We start by discussing its motivational issues, defining the author name disambiguation task and presenting its foundations. Next, we describe some methods proposed by our research group, as well as some recent approaches to author name disambiguation. Finally, we discuss open issues.
- 25 September 2023
- Published by Sociedade Brasileira de Computação
- p. 179-182
Open AccessAutomatic Disambiguation of Author Names: Foundations, Methods and Open Issues
This tutorial is based on our book “Automatic Disambiguation of Author Names in Bibliographic Repositories” and aims to spread the problem and its challenges among the SBBD community. Author name ambiguity problem occurs when an author publishes works under distinct names or distinct authors publish works under similar names. This problem may be caused by a number of reasons, including the lack of standards and common practices, and the decentralized generation of bibliographic content. In this tutorial, we intend to present an ample view on the automatic disambiguation of author names. We start by discussing its motivational issues, defining the author name disambiguation task and presenting its foundations. Next, we describe some methods proposed by our research group, as well as some recent approaches to author name disambiguation. Finally, we discuss open issues. This tutorial is based on our book “Automatic Disambiguation of Author Names in Bibliographic Repositories” and aims to spread the problem and its challenges among the SBBD community. Author name ambiguity problem occurs when an author publishes works under distinct names or distinct authors publish works under similar names. This problem may be caused by a number of reasons, including the lack of standards and common practices, and the decentralized generation of bibliographic content. In this tutorial, we intend to present an ample view on the automatic disambiguation of author names. We start by discussing its motivational issues, defining the author name disambiguation task and presenting its foundations. Next, we describe some methods proposed by our research group, as well as some recent approaches to author name disambiguation. Finally, we discuss open issues.
- 16 August 2021
- Published by Springer Nature
- p. 229-243
Experimental Evaluation Among Reblocking Techniques Applied to the Entity Resolution
- 16 August 2021
- Published by Springer Nature
- p. 229-243
Experimental Evaluation Among Reblocking Techniques Applied to the Entity Resolution
- Open Access
- 29 October 2019
- Published by Association for Computing Machinery (ACM)
- p. 425-432
An item-item similarity approach based on linked open data semantic relationship
- 29 October 2019
- Published by Association for Computing Machinery (ACM)
- p. 425-432
Open AccessAn item-item similarity approach based on linked open data semantic relationship
- Open Access
- 25 August 2018
- Published by Sociedade Brasileira de Computação
- p. 61-72
Melhorias no Processo de Blocagem para Resolução de Entidades Baseadas na Relevância dos Termos
Entity Resolution is a task commonly faced in data integration process. Due to quadratic number of comparisons to decide those instances belonging to the same entity, we need another way for performing such comparisons. In order to mitigate such a problem, techniques of blocking and block processing have been applied aiming the efficiency. In this work, we propose options to choose terms in the blocking step based on their relevance to the dataset in the phases of blocking and processing of blocks. We assess our proposal comparing it against relevant works available in the literature. The results show that our proposal decrease the run time by half, increasing the efficiency. Resolução de Entidades é uma tarefa comumente enfrentada no processo de integração de dados. Por necessitar de um número de comparações de ordem quadrática, torna-se inviável aplica-lá em grandes conjuntos de dados. Técnicas de blocagem e de processamentos de blocos têm sido propostas, visando amenizar esse problema. Neste trabalho, é proposta uma forma de escolher termos para serem usados na etapa de blocagem e no processamento de blocos, com base em sua relevância na coleção de dados. A proposta é avaliada, comparando-a com trabalhos relevantes publicados na literatura. Os resultados mostram que a proposta deste trabalho reduz o tempo de processamento pela metade e melhora a qualidade dos blocos gerados.
- 25 August 2018
- Published by Sociedade Brasileira de Computação
- p. 61-72
Open AccessMelhorias no Processo de Blocagem para Resolução de Entidades Baseadas na Relevância dos Termos
Entity Resolution is a task commonly faced in data integration process. Due to quadratic number of comparisons to decide those instances belonging to the same entity, we need another way for performing such comparisons. In order to mitigate such a problem, techniques of blocking and block processing have been applied aiming the efficiency. In this work, we propose options to choose terms in the blocking step based on their relevance to the dataset in the phases of blocking and processing of blocks. We assess our proposal comparing it against relevant works available in the literature. The results show that our proposal decrease the run time by half, increasing the efficiency. Resolução de Entidades é uma tarefa comumente enfrentada no processo de integração de dados. Por necessitar de um número de comparações de ordem quadrática, torna-se inviável aplica-lá em grandes conjuntos de dados. Técnicas de blocagem e de processamentos de blocos têm sido propostas, visando amenizar esse problema. Neste trabalho, é proposta uma forma de escolher termos para serem usados na etapa de blocagem e no processamento de blocos, com base em sua relevância na coleção de dados. A proposta é avaliada, comparando-a com trabalhos relevantes publicados na literatura. Os resultados mostram que a proposta deste trabalho reduz o tempo de processamento pela metade e melhora a qualidade dos blocos gerados.
- 1 September 2017
- Published by Elsevier in Information Processing & Management
- Vol. 53 (5) , 1120-1138
DERIN: A data extraction method based on rendering information and n-gram
- 1 September 2017
- Published by Elsevier in Information Processing & Management
- Vol. 53 (5) , 1120-1138
DERIN: A data extraction method based on rendering information and n-gram
- 1 September 2014
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 173-182
Combining domain-specific heuristics for author name disambiguation
Author name disambiguation has been one of the hardest problems faced by digital libraries since their early days. Historically, supervised solutions have empirically outperformed those based on heuristics, but with the burden of having to rely on manually labelled training sets for the learning process. Moreover, most supervised solutions just apply some type of generic machine learning solution and do not exploit specific knowledge about the problem. In this paper, we follow a similar reasoning, but in the opposite direction. Instead of extending an existing supervised solution, we propose a set of carefully designed heuristics and similarity functions and apply supervision only to optimize such parameters for each particular dataset. As our experiments show, the result is a very effective, efficient and practical author name disambiguation method that can be used in many different scenarios.
- 1 September 2014
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 173-182
Combining domain-specific heuristics for author name disambiguation
Author name disambiguation has been one of the hardest problems faced by digital libraries since their early days. Historically, supervised solutions have empirically outperformed those based on heuristics, but with the burden of having to rely on manually labelled training sets for the learning process. Moreover, most supervised solutions just apply some type of generic machine learning solution and do not exploit specific knowledge about the problem. In this paper, we follow a similar reasoning, but in the opposite direction. Instead of extending an existing supervised solution, we propose a set of carefully designed heuristics and similarity functions and apply supervision only to optimize such parameters for each particular dataset. As our experiments show, the result is a very effective, efficient and practical author name disambiguation method that can be used in many different scenarios.