Chang Sun
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Chang Sun has a total of 36 co-authors. Here are the top co-authors by number of shared publications:
- Johan van Soest
Brightlands Institute for Smart Society (BISS), Faculty of Science and Engineering, Maastricht University, The Netherlands
4 shared publications - Michel Dumontier
Institute of Data Science, Maastricht University, Maastricht, Netherlands
4 shared publications Michel Dumontier
Maastricht University
3 shared publications- Andre Dekker
Department of Radiation Oncology (Maastro), GROW Research Institute for Oncology and Reproduction, Maastricht University Medical Centre+, Maastricht, The Netherlands
2 shared publications - Annemarie Koster
Department of Social Medicine, CAPHRI Care and Public Health Research Institute, Maastricht University, Maastricht, The Netherlands
2 shared publications - Lianne Ippel
Tilburg University
2 shared publications
Latest Publications
- Open Access
- 18 September 2026
- Published by JMIR Publications Inc. in JMIR Preprints
Patient Care Advocate for Selectively Proactive Information Support During Clinical Consultations: Conceptual Framework (Preprint)
UNSTRUCTURED During consultations, clinicians must make decisions while relevant patient history, preferences, medical evidence, and outputs from a growing range of decision-support technologies remain dispersed across sources and difficult to integrate when needed. The challenge is not only information availability, but identifying what is relevant to the evolving discussion and introducing it at the appropriate moment without creating unnecessary distraction. This Viewpoint introduces PCAdvocate as a patient-oriented information-support role and proposes a design pattern governed by the principle of selective proactivity. To realize this principle, we define five core capabilities of PCAdvocate as its functional requirements–consultation awareness, longitudinal patient-context access, information-need detection, targeted information seeking, and relevance judgment and calibrated information delivery–and integrate them into a functional framework showing how they work together. We then illustrate this against existing clinical technologies, finding individual capabilities are variably supported across technology classes, but none integrates all five within a system. Finally, we illustrate how the framework could be operationalized through system configuration and use-case scenario, and outline strategies for evaluating future PCAdvocate systems.
- 18 September 2026
- Published by JMIR Publications Inc. in JMIR Preprints
Open AccessPatient Care Advocate for Selectively Proactive Information Support During Clinical Consultations: Conceptual Framework (Preprint)
UNSTRUCTURED During consultations, clinicians must make decisions while relevant patient history, preferences, medical evidence, and outputs from a growing range of decision-support technologies remain dispersed across sources and difficult to integrate when needed. The challenge is not only information availability, but identifying what is relevant to the evolving discussion and introducing it at the appropriate moment without creating unnecessary distraction. This Viewpoint introduces PCAdvocate as a patient-oriented information-support role and proposes a design pattern governed by the principle of selective proactivity. To realize this principle, we define five core capabilities of PCAdvocate as its functional requirements–consultation awareness, longitudinal patient-context access, information-need detection, targeted information seeking, and relevance judgment and calibrated information delivery–and integrate them into a functional framework showing how they work together. We then illustrate this against existing clinical technologies, finding individual capabilities are variably supported across technology classes, but none integrates all five within a system. Finally, we illustrate how the framework could be operationalized through system configuration and use-case scenario, and outline strategies for evaluating future PCAdvocate systems.
- Open Access
- 1 February 2026
- Published by SAGE Publications in Data Science
- Vol. 9 (1)
Knowledge Graph Metadata Specification and Validation
Knowledge graphs (KGs) are published across diverse repositories using heterogeneous metadata models that vary in their descriptive elements, level of detail, vocabularies, and support for semantic interoperability, in part due to the absence of a standardized specification. This lack of consistency in metadata practices hinders effective discovery and reuse of KGs. To address this gap, we present a KG metadata specification and validation framework designed to foster structured, FAIR-aligned KG descriptions. Developed through a community-driven process, the specification defines 33 metadata elements formalized using the Shapes Constraint Language (SHACL) to enable automated validation. We applied the framework to metadata from 1,573 datasets in the Linked Open Data Cloud by mapping its 17 metadata fields to those in our specification. This mapping demonstrated the specification’s compatibility with the largest Linked Data repository and validation process revealed common issues and areas for improvement in metadata quality. Furthermore, we demonstrated how metadata can be published in a way that enhances its visibility for end users and its discoverability by search engines, achieving full compliance with FAIR-Checker evaluations through the publication of specification-conformant metadata for the Food Health Claims KG. The framework provides a foundation for KG publishers, repository managers, and researchers, promoting uniform, high-quality metadata practices and advancing FAIR principles within the KG community.
- 1 February 2026
- Published by SAGE Publications in Data Science
- Vol. 9 (1)
Open AccessKnowledge Graph Metadata Specification and Validation
Knowledge graphs (KGs) are published across diverse repositories using heterogeneous metadata models that vary in their descriptive elements, level of detail, vocabularies, and support for semantic interoperability, in part due to the absence of a standardized specification. This lack of consistency in metadata practices hinders effective discovery and reuse of KGs. To address this gap, we present a KG metadata specification and validation framework designed to foster structured, FAIR-aligned KG descriptions. Developed through a community-driven process, the specification defines 33 metadata elements formalized using the Shapes Constraint Language (SHACL) to enable automated validation. We applied the framework to metadata from 1,573 datasets in the Linked Open Data Cloud by mapping its 17 metadata fields to those in our specification. This mapping demonstrated the specification’s compatibility with the largest Linked Data repository and validation process revealed common issues and areas for improvement in metadata quality. Furthermore, we demonstrated how metadata can be published in a way that enhances its visibility for end users and its discoverability by search engines, achieving full compliance with FAIR-Checker evaluations through the publication of specification-conformant metadata for the Food Health Claims KG. The framework provides a foundation for KG publishers, repository managers, and researchers, promoting uniform, high-quality metadata practices and advancing FAIR principles within the KG community.
- Open Access
Adapting Lightweight Vision Language Models for Radiological Visual Question Answering
Recent advancements in vision-language systems have improved the accuracy of Radiological Visual Question Answering (VQA) Models. However, some challenges remain across each stage of model development: limited expert-labeled images hinders data procurement at scale; the intricate and nuanced patterns of radiological images make modeling inherently difficult; and the lack of evaluation evaluation efforts makes it difficult to identify cases where the model might be ill-conditioned. In this study, we fine-tune a lightweight 3B parameter vision-language model for Radiological VQA, demonstrating that small models, when appropriately tuned with curated data, can achieve robust performance across both open- and closed-ended questions. We propose a cost-effective training pipeline from synthetic question-answer pair generation to multi-stage fine-tuning on specialised radiological domain-targeted datasets (e.g., ROCO v2.0, MedPix v2.0). Our results show that despite operating at a fraction of the scale of state-of-the-art models such as LLaVA-Med, our model achieves promising performance given its small parameter size and the limited scale of training data. We introduce a lightweight saliency-based diagnostic tool that enables domain experts to inspect VQA model performance and identify ill-conditioned failure modes through saliency analysis.
Open AccessAdapting Lightweight Vision Language Models for Radiological Visual Question Answering
Recent advancements in vision-language systems have improved the accuracy of Radiological Visual Question Answering (VQA) Models. However, some challenges remain across each stage of model development: limited expert-labeled images hinders data procurement at scale; the intricate and nuanced patterns of radiological images make modeling inherently difficult; and the lack of evaluation evaluation efforts makes it difficult to identify cases where the model might be ill-conditioned. In this study, we fine-tune a lightweight 3B parameter vision-language model for Radiological VQA, demonstrating that small models, when appropriately tuned with curated data, can achieve robust performance across both open- and closed-ended questions. We propose a cost-effective training pipeline from synthetic question-answer pair generation to multi-stage fine-tuning on specialised radiological domain-targeted datasets (e.g., ROCO v2.0, MedPix v2.0). Our results show that despite operating at a fraction of the scale of state-of-the-art models such as LLaVA-Med, our model achieves promising performance given its small parameter size and the limited scale of training data. We introduce a lightweight saliency-based diagnostic tool that enables domain experts to inspect VQA model performance and identify ill-conditioned failure modes through saliency analysis.
- Open Access
- 3 January 2025
- Published by Springer Nature in npj Digital Medicine
- Vol. 8 (1) , 1-14
Generating unseen diseases patient data using ontology enhanced generative adversarial networks
Generating realistic synthetic health data (e.g., electronic health records), holds promise for fundamental research, AI model development, and enhancing data privacy safeguards. Generative Adversarial Networks (GANs) have been employed for this purpose, but their performance is largely constrained by their reliance on training data, rendering them inadequate for rare or previously unseen diseases. This study proposes Onto-CGAN, a novel generative framework that combines knowledge from disease ontologies with GANs to generate unseen diseases that are not present in the training data. The quality of the generated data is evaluated using variable distributions, correlation coefficients, and machine learning model performance. Our findings demonstrate that Onto-CGAN generates unseen diseases with statistical characteristics comparable to the real data, and significantly improves the training of machine learning models. This innovative approach addresses the scarcity of data for rare diseases, offering valuable applications in data augmentation, hypothesis generation, and preclinical validation of clinical models.
- 3 January 2025
- Published by Springer Nature in npj Digital Medicine
- Vol. 8 (1) , 1-14
Open AccessGenerating unseen diseases patient data using ontology enhanced generative adversarial networks
Generating realistic synthetic health data (e.g., electronic health records), holds promise for fundamental research, AI model development, and enhancing data privacy safeguards. Generative Adversarial Networks (GANs) have been employed for this purpose, but their performance is largely constrained by their reliance on training data, rendering them inadequate for rare or previously unseen diseases. This study proposes Onto-CGAN, a novel generative framework that combines knowledge from disease ontologies with GANs to generate unseen diseases that are not present in the training data. The quality of the generated data is evaluated using variable distributions, correlation coefficients, and machine learning model performance. Our findings demonstrate that Onto-CGAN generates unseen diseases with statistical characteristics comparable to the real data, and significantly improves the training of machine learning models. This innovative approach addresses the scarcity of data for rare diseases, offering valuable applications in data augmentation, hypothesis generation, and preclinical validation of clinical models.
- Open Access
- 1 January 2025
- Published by Association for Computational Linguistics (ACL)
- p. 54-67
Machine Unlearning of Personally Identifiable Information in Large Language Models
Dan Parii, Thomas van Osch, Chang Sun. Proceedings of the Natural Legal Language Processing Workshop 2025. 2025.
- 1 January 2025
- Published by Association for Computational Linguistics (ACL)
- p. 54-67
Open AccessMachine Unlearning of Personally Identifiable Information in Large Language Models
Dan Parii, Thomas van Osch, Chang Sun. Proceedings of the Natural Legal Language Processing Workshop 2025. 2025.
- Open Access
- 1 January 2025
- Published by Association for Computational Linguistics (ACL)
- p. 125-134
Copyright Infringement by Large Language Models in the EU: Misalignment, Safeguards, and the Path Forward
Noah Scharrenberg, Chang Sun. Proceedings of the Natural Legal Language Processing Workshop 2025. 2025.
- 1 January 2025
- Published by Association for Computational Linguistics (ACL)
- p. 125-134
Open AccessCopyright Infringement by Large Language Models in the EU: Misalignment, Safeguards, and the Path Forward
Noah Scharrenberg, Chang Sun. Proceedings of the Natural Legal Language Processing Workshop 2025. 2025.
- Open Access
Empirical Privacy Evaluations of Generative and Predictive Machine Learning Models -- A review and challenges for practice
Synthetic data generators, when trained using privacy-preserving techniques like differential privacy, promise to produce synthetic data with formal privacy guarantees, facilitating the sharing of sensitive data. However, it is crucial to empirically assess the privacy risks associated with the generated synthetic data before deploying generative technologies. This paper outlines the key concepts and assumptions underlying empirical privacy evaluation in machine learning-based generative and predictive models. Then, this paper explores the practical challenges for privacy evaluations of generative models for use cases with millions of training records, such as data from statistical agencies and healthcare providers. Our findings indicate that methods designed to verify the correct operation of the training algorithm are effective for large datasets, but they often assume an adversary that is unrealistic in many scenarios. Based on the findings, we highlight a crucial trade-off between the computational feasibility of the evaluation and the level of realism of the assumed threat model. Finally, we conclude with ideas and suggestions for future research.
Open AccessEmpirical Privacy Evaluations of Generative and Predictive Machine Learning Models -- A review and challenges for practice
Synthetic data generators, when trained using privacy-preserving techniques like differential privacy, promise to produce synthetic data with formal privacy guarantees, facilitating the sharing of sensitive data. However, it is crucial to empirically assess the privacy risks associated with the generated synthetic data before deploying generative technologies. This paper outlines the key concepts and assumptions underlying empirical privacy evaluation in machine learning-based generative and predictive models. Then, this paper explores the practical challenges for privacy evaluations of generative models for use cases with millions of training records, such as data from statistical agencies and healthcare providers. Our findings indicate that methods designed to verify the correct operation of the training algorithm are effective for large datasets, but they often assume an adversary that is unrealistic in many scenarios. Based on the findings, we highlight a crucial trade-off between the computational feasibility of the evaluation and the level of realism of the assumed threat model. Finally, we conclude with ideas and suggestions for future research.
- 31 May 2024
- Published by Springer Nature
- p. 389-399
Mitigating Class Imbalance in Time Series with Enhanced Diffusion Models
- 31 May 2024
- Published by Springer Nature
- p. 389-399
Mitigating Class Imbalance in Time Series with Enhanced Diffusion Models
- Open Access
- 1 June 2023
- Published by Elsevier in Journal of Biomedical Informatics
- Vol. 143, 104404
Generating synthetic personal health data using conditional generative adversarial networks combining with differential privacy
- 1 June 2023
- Published by Elsevier in Journal of Biomedical Informatics
- Vol. 143, 104404
Open AccessGenerating synthetic personal health data using conditional generative adversarial networks combining with differential privacy
- Open Access
- 10 May 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 11, 47304-47320
Deep Generative Models for Synthetic Data: A Survey
A growing interest in synthetic data has stimulated the development and advancement of a large variety of deep generative models for a wide range of applications. However, as this research has progressed, its streams have become more specialized and disconnected from one another. This is why models for synthesizing text data for natural language processing cannot readily be compared to models for synthesizing health records anymore. To mitigate this isolation, we propose a data-driven evaluation framework for generative models for synthetic sequential data, an important and challenging sub-category of synthetic data, based on five high-level criteria: representativeness, novelty, realism, diversity and coherence of a synthetic data-set relative to the original data-set regardless of the models’ internal structures. The criteria reflect requirements different domains impose on synthetic data and allow model users to assess the quality of synthetic data across models. In a critical review of generative models for sequential data, we examine and compare the importance of each performance criterion in numerous domains. We find that realism and coherence are more important for synthetic data natural language, speech and audio processing tasks. At the same time, novelty and representativeness are more important for healthcare and mobility data. We also find that measurement of representativeness is often accomplished using statistical metrics, realism by using human judgement, and novelty using privacy tests.
- 10 May 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 11, 47304-47320
Open AccessDeep Generative Models for Synthetic Data: A Survey
A growing interest in synthetic data has stimulated the development and advancement of a large variety of deep generative models for a wide range of applications. However, as this research has progressed, its streams have become more specialized and disconnected from one another. This is why models for synthesizing text data for natural language processing cannot readily be compared to models for synthesizing health records anymore. To mitigate this isolation, we propose a data-driven evaluation framework for generative models for synthetic sequential data, an important and challenging sub-category of synthetic data, based on five high-level criteria: representativeness, novelty, realism, diversity and coherence of a synthetic data-set relative to the original data-set regardless of the models’ internal structures. The criteria reflect requirements different domains impose on synthetic data and allow model users to assess the quality of synthetic data across models. In a critical review of generative models for sequential data, we examine and compare the importance of each performance criterion in numerous domains. We find that realism and coherence are more important for synthetic data natural language, speech and audio processing tasks. At the same time, novelty and representativeness are more important for healthcare and mobility data. We also find that measurement of representativeness is often accomplished using statistical metrics, realism by using human judgement, and novelty using privacy tests.