https://repositorio.ufba.br/handle/ri/34151 2026-04-17T05:57:03Z 2026-04-17T05:57:03Z Santos, Juliana Conceição https://repositorio.ufba.br/handle/ri/44334 2026-04-06T23:34:58Z 2025-06-05T00:00:00Z

Process checklist: checklist orientado à transparência para inspeção de processos BPMN Santos, Juliana Conceição Maciel, Rita Suzana Pitangueira In recent years there has been an interest in quality assurance processes for Process models in BPMN. This can be achieved through inspection, a static analysis technique that demonstrates potential for identifying problems in software artifacts. Inspection of process models with checklists, although still little explored in the process literature, is an important instrument to support the inspection process in detecting defects, aiming at the quality of the artifacts. The complexity of process models, the lack of studies on inspection performed by humans and the search for quality of the models have driven this research, focusing on quality Transparency. The transparency of BPMN models contributes to the description of process models, improving understanding and comprehension, benefiting not only internal efficiency such as management and communication, but also resulting in strategic benefits, such as good reputation, credibility and the sharing of quality information. Thus, by evaluating the items of the checklist that supports the assessment of the quality of the models, systematically and adding the knowledge of the Transparency Catalog, in the checklist of an organization in the justice sector, opportunities were identified to evolve the practices of verification of process models in BPMN. This master's dissertation aims to propose a Checklist (Process Checklist), an inspection tool focused on transparency, to enhance the quality of process models in BPMN, identifying quality problems in the model. For this purpose, a literature review was initially carried out to identify studies on inspection of BPMN process models. The checklist, developed for human use, was evaluated anonymously by process modeling experts. Based on the results of the first evaluation of the usability, efficiency and effectiveness of the Process Checklist, carried out by five experts, opportunities for improvement were identified that were implemented to ensure transparent, objective, reliable and quality process models. In the second performance evaluation, the Process Checklist was more effective in ensuring model quality compared to BPCheck. Universidade Federal da Bahia Dissertação

2025-06-05T00:00:00Z Proencia, Edlane Cristine dos Santos https://repositorio.ufba.br/handle/ri/44167 2026-03-03T10:32:16Z 2025-12-15T00:00:00Z

Interoperabilidade organizacional no contexto dos sistemas de sistemas de informação: uma abordagem colaborativa. Proencia, Edlane Cristine dos Santos Maciel, Rita Suzana Pitangueira The interaction among independent, heterogeneous, and dynamic systems in Systems of Information Systems (SoIS) imposes significant challenges, especially at the level of Organizational Interoperability. To achieve this interoperability, aligning business processes across organizations is fundamental. Business Process Management (BPM) approaches are essential tools for this alignment, but existing solutions fail in collaborative contexts where participants can join and leave at any time, without central management. Given this gap, the main objective of this work was to develop a BPM-based solution to support Organizational Interoperability in collaborative SoIS. The main contribution was the specification of a metamodel that provides a management structure. This metamodel enables modeling collaborative business processes and addresses the entry and exit dynamics of participants by formalizing the explicit separation between Role (the abstract function responsible for the process) and Participant (the concrete and dynamic organizational instance that executes it). Additionally, the solution allows the definition of the Interoperability Links (technical and sociotechnical) required for interactions and communications between roles, ensuring that collaboration requirements are explicit. The solution’s applicability was evaluated through its implementation in a semi-automated simulation environment instantiated in three domains (e-commerce, health plan, and public security). The results validated the approach, demonstrating adaptability upon the entry of new participants and resilience upon exit, through Role-based rerouting. It is concluded that the solution offers a viable model to manage participant dynamics, providing the transparency and resilience mechanisms necessary for Organizational Interoperability in collaborative SoIS. Universidade Federal da Bahia Dissertação

2025-12-15T00:00:00Z Joel, Machado Pires https://repositorio.ufba.br/handle/ri/44141 2026-03-02T12:08:30Z 2026-01-20T00:00:00Z

Integrating confidence into embedding-based models for learn-to-rank in recommender systems. Joel, Machado Pires Durão, Frederico Araújo Recommender systems (RecSys) enhance information retrieval efficiency across different domains by delivering personalized content that aligns with user preferences. These systems address user-item relationships through methods like matrix factorization and graph attention networks (GAT). Despite advancements in accuracy, existing approaches often narrowly focus on predictive performance while neglecting the broader utility of confidence estimation. This estimation is crucial for quantifying the certainty behind recommendations, particularly in situations where a balance between risk and reward is required. RecSys can mitigate uncertainties stemming from data noise and model limitations by utilizing confidence. Existing approaches to confidence integration face critical limitations. Non-parametric techniques, such as neural network-based probabilistic calibration, remain confined to classification tasks, failing to address regression scenarios, including rating prediction and listwise learn-to-rank. Many confidence models operate independently of core recommendation processes, limiting their adaptability and calibration impact. Notably, the literature neglects the integration of confidence into GAT-based models. Additionally, the literature lacks experimental evaluation of different distribution-based methods. Therefore, this study proposes an experimental evaluation of previous distribution-based methods and explores a suitable confidence integration in GAT-based models. We evaluate four prior solutions in terms of rating prediction accuracy, ranking accuracy, and confidence correlation with error. These solutions and our proposal are evaluated in public datasets from varying contexts and characteristics. Results reveal that distribution-based confidence integration often harms models' accuracy and leaves room for improvement regarding the correlation between confidence and error. Although these findings also hold for our method, it still achieves superior performance compared to all prior solutions and shows promising results in terms of negative confidence–error correlation. Furthermore, as a second part of this study, we propose and evaluate the integration of confidence into embedding models for learn-to-rank methods. This proposal and its baselines are also evaluated across various public datasets, using different ranking metrics, and the correlation with confidence and error. The results reveal that both proposed methods consistently demonstrate competitive rank performances and even outperform the baselines in some datasets. Specifically, the proposed confidence integration for rating prediction achieved improvements of at least 58.16% in ranking metrics on the Amazon Movies and TVs, 34.94% on the Jester Joke, and 42.98% on the MovieLens dataset. Additionally, we observed a cubic polynomial relationship between confidence and error in this latter solution. Universidade Federal da Bahia Dissertação

2026-01-20T00:00:00Z Sant'Ana Filho, Marcos Vinícius Queiroz de https://repositorio.ufba.br/handle/ri/44119 2026-02-27T10:36:29Z 2025-11-27T00:00:00Z

Paperman - um sistema de recomendação de artigos científicos. Sant'Ana Filho, Marcos Vinícius Queiroz de Durão, Frederico Araújo The search for references and related work in scientific research can be an exhausting process, consuming an average of 4 hours per week for researchers. The abundance of sources and repositories makes it even more challenging to verify the veracity and reliability of these works, hindering the selection of high-quality, relevant research and leading to the disposal of half of the collected samples, negatively impacting productivity. Considering this scenario, the objective of this study is to plan and develop a platform to facilitate the initial stages of research through recommendation systems, models based on the researcher's profile, and data post-processing. The proposed system, called Paperman, employs natural language processing and machine learning techniques to analyze researchers' publication history and generate personalized recommendations for scientific articles. The system architecture includes an API for data collection and processing, integrations with external services such as ORCID and DBLP, and a browser extension that presents recommendations intuitively. Experimental results demonstrate the system's effectiveness, with metrics such as MRR of 0.8 and nDCG@5 of 0.9407, indicating high relevance of generated recommendations. The study contributes to the field of educational recommendation systems, offering a practical solution to optimize the literature review process and discovery of related works in scientific research. The Paperman system addresses common challenges in academic research, such as information overload and the need for efficient discovery of relevant publications, by leveraging the researcher's own profile and publication history to provide tailored recommendations. Universidade Federal da Bahia Dissertação

2025-11-27T00:00:00Z