Los ingenieros de requisitos deben adquirir información acerca del contexto don-de se desempeñará el futuro sistema de software y registrarla en modelos apro-piados. Estos modelos son utilizados durante la definición de los requisitos del sistema de software, pero también sirven de referencia para atender cualquier du-da que pueda surgir en etapas posteriores del desarrollo. Desafortunadamente, se ha comprobado que estos modelos dependen muy fuertemente de quien los cons-truya. Esto hace necesario que procedimientos y recomendaciones que acompa-ñan a los modelos consideren factores cognitivos, de manera que los resultados no sean tan dependientes de las personas. En tal sentido, se ha reformulado un proceso de construcción de un modelo en lenguaje natural con heurísticas que atienden dichos aspectos cognitivos y semánticos. Se presenta un experimento que utiliza este proceso modificado y se lo compara con experimentos antecesores donde se usó un proceso sin consideraciones lingüísticas-cognitivas y un proceso que atiende solo algunas cuestiones lingüísticas para construir el mismo modelo. Se pudieron evidenciar mejoras en la calidad del modelo cuando se utiliza un pro-ceso con heurísticas más precisas que atiende tanto aspectos semánticos y prag-máticos como cognitivos.
La calidad de los modelos depende no sólo de las habilidades de los ingenieros de requisitos y del proceso de construcción, sino de la información adquirida. Muchos defectos en los modelos se originan en información deficiente de la elicitación, habitualmente de entrevistas. Esto revela dificultades en la comprensión del universo de discurso, del problema del usuario y de la funcionalidad general del software. Este trabajo propone mejorar esa comprensión utilizando técnicas visuales, como mapas mentales y facilitación gráfica. Se presentan los resultados de un análisis comparativo de las percepciones de estudiantes actuando como ingenieros junior de requisitos, a partir de encuestas realizadas en dos estudios de caso sobre el uso de estas técnicas durante la actividad de elicitación. Con ambas técnicas se construyeron diagramas individualmente, y luego se unificaron. Esto resultó ser clave para contrastar discrepancias e integrar diversos puntos de vista, así como base para construir un modelo conceptual. Los resultados indican que el nivel de comprensión alcanzado fue percibido como alto para ambas técnicas. Además, ambas facilitaron la interacción con el usuario. Se identificaron ciertas limitaciones, tales como la necesidad de adquirir experiencia en el uso de estas técnicas, y la importancia de realizar alguna reunión para discutir y validar el conocimiento adquirido mediante los diagramas.
Requirements Engineering (RE) is a critical phase in software development. Frequently, requirements are expressed in natural language and embedded in large documents. Classifying requirements is a time-consuming and error-prone activity. To address this issue, Machine Learning (ML) techniques can be leveraged to classify requirements. ML is a subfield of Artificial Intelligence that facilitates decision-making by developing automated models trained on data samples. This paper compares ML approaches to automatically classify requirements as functional (FR) and non-functional (NFR). Our study uses Supervised Machine Learning (SL) models alongside Active Learning (AL). SL models require large volumes of labeled data for effective training. However, in many cases, the datasets available for training are unlabeled, and the sheer size of these datasets makes manual labeling impractical. This poses a significant challenge for training supervised models. AL offers a solution to this challenge by strategically selecting specific xamples for user labeling, which are then used to train the model. By combining AL with SL, we aim to investigate whether Active Learning improves requirements classification. We propose a systematic approach to applying ML and AL to classify requirements datasets. By doing so, we aim to accelerate and automate the requirements classification process. Classifying requirements into distinct categories allows developers to concentrate more effectively on subsequent stages of the development process. We present a supporting process for our proposal.
Recent advancements in artificial intelligence (AI), particularly in large language models (LLMs), offer new possibilities for automating requirements generation from elicitation interviews. This study compares the performance of ChatGPT-4 and DeepSeek-V3 in generating software requirements based on transcribed stakeholder interviews. Using two case studies, the LLMs were tasked with identifying functional and non-functional requirements. The results indicate that ChatGPT-4 performed better in extracting precise requirements, particularly nonfunctional ones, while DeepSeek-V3 demonstrated advantages in efficiency. However, both models exhibited limitations in handling ambiguity and properly categorizing requirements. This study highlights the potential of LLMs in Requirements Engineering while emphasizing the need for improved prompt/dialogues techniques and human supervision. Future research should explore hybrid AI-human approaches and domain-specific fine-tuning to enhance requirement extraction accuracy.
User stories are one of the most widely used artifacts in the software industry to define functional requirements. In parallel, the use of high-fidelity mockups facilitates end-user participation in defining their needs. In this work, we explore how combining these techniques with large language models (LLMs) enables agile and automated generation of user stories from mockups. To this end, we present a case study that analyzes the ability of LLMs to extract user stories from high-fidelity mockups, both with and without the inclusion of a glossary of the Language Extended Lexicon (LEL) in the prompts. Our results demonstrate that incorporating the LEL significantly enhances the accuracy and suitability of the generated user stories. This approach represents a step forward in the integration of AI into requirements engineering, with the potential to improve communication between users and developers.
A Engenharia de Requisitos ainda enfrenta desafios significativos na elicitação e especificação de requisitos de usabilidade, especialmente quando conduzida por Engenheiros de Software em formação. O Método USARP (USAbility Requirements with Personas and user stories) tem se mostrado promissor nesse contexto, ao incorporar a perspectiva do usuário desde as fases iniciais do desenvolvimento. Entretanto, o corpo de conhecimento da área carece de padrões formalizados voltados à usabilidade, dificultando a padronização e o reúso de requisitos. Diante dessa lacuna, este estudo propõe a análise de requisitos especificados com o uso do USARP, com o objetivo de identificar padrões emergentes. A investigação foi realizada com base em 24 projetos acadêmicos desenvolvidos por estudantes de Engenharia de Software, uma vez que o acesso a requisitos de projetos industriais é restrito por questões de confidencialidade. Como resultado, foram identificados 12 padrões recorrentes relacionados às funcionalidades: Cadastro, Busca, Exclusão, Login, Filtragem, Edição, Envio de Arquivos, Acompanhamento de Progresso, Geração de Relatórios, Gerenciamento, Personalização e Notificação. Esses padrões contribuem para a sistematização do método USARP, promovendo o reúso de requisitos e a melhoria da qualidade dos artefatos produzidos.
La fase de requisitos es importante para el éxito de un proyecto de software, ya que un levantamiento bien definido evita costos altos por correcciones y asegura productos alineados con las expectativas del usuario. Sin embargo, el enfoque tradicional suele centrarse en requisitos funcionales y no funcionales, ignorando la interacción del usuario, lo que genera interfaces (UI) poco usables. Este trabajo propone un nuevo enfoque que incorpora patrones de interacción validados en la fase de requisitos para optimizar el diseño de UI. La metodología consta de tres etapas: 1) identificación sistemática de patrones de interacción relevantes para el dominio, 2) análisis contextual de requisitos y 3) mapeo de patrones a requisitos mediante validación heurística. Se evaluó mediante un caso de estudio en una aplicación web para actualización de datos con validación biométrica, con doce requisitos implementados por un equipo de cinco desarrolladores. Los resultados preliminares muestran que los patrones optimizaron la estructura de la UI y la retroalimentación, alineando requisitos con estándares de usabilidad. Los resultados sugieren que el enfoque no solo alinea el diseño con mejores prácticas, sino que también ofrece beneficios medibles en eficiencia y mantenibilidad. Futuros estudios podrían cuantificar su impacto con métricas como SUS y explorar su escalabilidad en sistemas complejos, integrar sistemáticamente las pautas de accesibilidad WCAG, y el uso de aprendizaje automático para recomendación adaptativa de patrones.
Due to safety and performance requirements, lithium-ion cell testing is critical in aerospace applications. Manual testing methods are error-prone and inefficient, leading to inconsistent results and poor traceability. This paper presents a requirements-driven methodology for automating Li-ion charger testing using LabVIEW. The approach utilizes the IEC/ISO/IEEE 29148-2018 writing requirements syntax to assist the software design, improving maintainability and reducing technical debt. A case study illustrates the development of an automated system for capturing charging curves, linking each code section to specific requirements. The solution improved debugging, documentation, and test coverage while reducing misunderstandings of informal specifications. Although direct performance comparisons are limited, automation allowed operators to focus on parallel tasks, enhancing man-hour efficiency. Results show that requirements traceability improved code clarity and streamlined communication between stakeholders. This structured method can be applied to other graphical programming environments where test coverage, maintainability, and resource optimization are essential.
The Softgoal Interdependency Graph (SIG) models non-functional requirements (NFRs) by representing softgoals and their interrelationships. However, building a SIG is challenging as it requires a deep understanding of qualitative concepts that vary across domains. The Transparency SIG (TSIG), which integrates over 30 related qualities, exemplifies this complexity. This study explores whether Large Language Models (LLMs), specifically ChatGPT-3.5 and ChatGPT-4o, can augment the knowledge of the TSIG. Through interactive dialogues, we analyzed the models' ability to suggest relevant content and structure. Our findings show that, using the TSIG as the Gold Standard, the ChatGPT-generated models demonstrated the ability to approximate the expert knowledge represented in the TSIG, as evidenced by three authors achieving over 84% recall. Furthermore, since precision varied significantly—from 29.4% to 100%—this highlights differences in the amount of false positives. These elements require further qualitative evaluation to determine which of them may actually contribute to augmenting the knowledge on transparency, as modeled by the TSIG.
Un Data Warehouse (almacén de datos) es un sistema de almacenamiento diseñado para recopilar, organizar y analizar grandes volúmenes de datos provenientes de diversas fuentes. Se utiliza en el análisis de negocios y la toma de decisiones, ya que permite consultar información histórica de manera eficiente. Un Data Warehouse ayuda a las empresas a gestionar, analizar y visualizar datos de manera eficiente, mejorando la toma de decisiones basada en información confiable y consolidada. Uno de los modelos para diseñar un Data Warehouse es el modelo multidimensional, que permite estructurar los datos de manera eficiente para análisis y reportes. Este modelo facilita la navegación y exploración de la información desde diferentes perspectivas, optimizando el rendimiento en consultas analíticas. Aunque constituye un punto de partida valioso, no siempre resulta sencillo obtener una especificación de requerimientos ordenada y consistente, que describa de manera integral toda la funcionalidad del modelo. Este artículo propone una herramienta que permite la derivación del modelo multidimensional utilizando diversas técnicas de procesamiento de lenguaje natural a partir del glosario del dominio LEL.
The complexity and interconnectivity of modern software systems make cybersecurity a key concern. Each new device connected to a network increases security risks, and the rise of cyber threats and attack sophistication requires strong security measures from the start of development. Agile development methodologies are widely adopted for their ability to improve efficiency and flexibility in software development. However, these methodologies often lack clear guidance on how to incorporate security practices effectively. This gap can result in vulnerabilities being exploited by attackers, compromising the security of sensitive systems and data. Based on this fact, balancing security with agility is essential in today's fast-paced digital landscape. Considering the application of cybersecurity requirements in agile development processes, since the beginning of the software development lifecycle, we propose the ASF Framework, developed using the Design Science Research approach. The framework evaluation focuses on its application to the OWASP ASVS cybersecurity standard and Scrum, using user stories with Acceptance Criteria and the Definition of Done to ensure clear and measurable development goals. This assessment was conducted in the context of a web platform that connects consumers and service providers, simplifying the process of offering and hiring services. The evaluation demonstrates the applicability of the ASF framework in a real-world scenario, and the results indicate that ASF effectively supported the identification of security requirements within an agile development context.
Os Large Language Models (LLMs), que são amplamente adotados em domínios como processamento de linguagem natural e geração de código, emergem como ferramentas promissoras para otimizar a Engenharia de Re-quisitos (ER). Por meio de uma Revisão Sistemática da Literatura (RSL), este estudo visa analisar sistematicamente como os LLMs têm sido aplicados na ER, explorando suas contribuições, benefícios, desafios e implicações éticas. Seguindo o protocolo estabelecido, foram analisados 50 estudos de várias bases bibliográficas. Os resultados revelam que os LLMs são empregados em todas as fases da ER (elicitação, análise, documentação e validação), destacando-se em automação de tarefas repetitivas, refinamento de re-quisitos e melhoria da comunicação entre stakeholders. Benefícios incluem ganhos de eficiência (redução de tempo/esforço) e maior precisão na docu-mentação. Entretanto, desafios persistem, como inconsistências na saída dos modelos, dependência de intervenção humana, dificuldades em domí-nios especializados e riscos de vazamento de dados. Preocupações éticas, como transparência e privacidade, são pouco exploradas na literatura, apon-tando lacunas críticas. Conclui-se que os LLMs transformam a ER, mas exi-gem frameworks robustos de validação e políticas éticas para equilibrar inovação e responsabilidade.
In recent decades, with the growth of new technologies, there has been an increase in the amount of personal data stored and also in unauthorized exposure, causing concern in society. To counteract this, governments have enacted laws that regulate the storage and use of data, such as the LGPD in Brazil. Software companies have faced challenges in operationalizing the principles of the LGPD due to limited technical knowledge about the law. This doctoral study proposes developing an approach for the elicitation and operationalization of LGPD requirements based on patterns. Its use can help the software industry to develop products that are compliant with the LGPD in a more efficient way.
This paper presents an approach for eliciting the capabilities of Cyber-Physical Production Systems (CPPS) through an extension of the Language Extended Lexicon (LEL), referred to as LEL-C. CPPS integrate hardware, software, and physical components—often in dynamic interaction with human and environmental factors—posing new challenges to requirements engineering. To address these, we adopt the capability-oriented perspective of e-CORE and propose leveraging LEL-C to systematically capture domain knowledge across digital and physical dimensions. The proposed extension incorporates additional attributes such as component type, system and physical location, interactions, data sources, and temporal constraints. We apply LEL-C to a wildfire detection and suppression system as a use case, showing how the method supports the structured identification of current system capabilities. It is planned to advance the validation of this prelimi-nary version of the LEL-C by developing a complete glossary for a CPSS. This will also provide improvements to the proposal. This foundational work enables a consistent strategy for capability elicitation in CPPS, con-tributing to the broader goal of aligning semantic models with both soft-ware and physical process properties.
A área do Direito está evoluindo substancialmente através da inclusão de softwares, especialmente os de Inteligência Artificial. Atualmente, os softwares estão sendo usados, por exemplo, para auxiliar a definição de sentenças judiciais e para apoiar o trabalho de advogados, em um mercado em franca expansão. No entanto, autoridades da área do Direito apontam a preocupação com diversos aspectos qualitativos, inerentes à legalidade, que não estão sendo observados com rigor suficiente na construção destes softwares. Trata-se de requisitos não funcionais, conhecidamente complexos de se implementar e comumente negligenciados. Neste artigo, abordamos o mapeamento de metas flexíveis a partir dos princípios legais presentes no Devido Processo Legal (DPL). Diversos elementos do DPL são qualitativos, e devem estar presentes durante um processo judicial, caso contrário, há o risco de se tornar inválido. Considerando a relevância do DPL e a necessidade de conhecimento específico do domínio do Direito para se estabelecer conhecimento para a especificação adequada de requisitos de software, neste artigo é proposto o Softgoal Interdependency Graph (SIG) para o DPL. O grafo foi construído a partir do conhecimento elicitado em fontes de informação com relevância reconhecida na área do Direito. Posteriormente foi enriquecido e validado a partir da visão de 7 especialistas que atuam em um Tribunal de Justiça do Rio de Janeiro.
Assuring safety and security from the earliest stages of development in critical IoT systems requires a clear understanding of the system's objectives, boundaries, and operational context. This paper presents SafeSecRETS, a software tool for agile project planning of critical IoT systems. Through a canvas-based approach, SafeSecRETS supports requirements engineers and stakeholders in project scope definition and system requirements elicitation. The tool features a collaborative pipeline with interconnected building blocks, fostering engagement among information, people, and decision-making. Moreover, it assists in identifying key elements of the critical IoT system, such as components, safety and security aspects, and potential risks, by preparing the requirements analysis and specification through methods based on STPA (Systems Theoretic Process Analysis). Built on a layered event-driven architecture, SafeSecRETS leverages modern, scalable technologies to provide a high-quality web application. We also demonstrate how the tool supports the planning of an automated insulin delivery system.
Requirements elicitation is the process through which engineers interact with information sources to acquire knowledge about a specific domain. In the early stages of software projects, it is uncommon for engineers to elicit non-functional requirements (NFRs) as first-class requirements, as this often demands time and articulation that stakeholders may not readily provide. Large Language Models (LLMs), such as ChatGPT, trained on massive textual datasets, offer a promising opportunity to support this process by generating coherent and context-relevant information about qualities that are key to a given domain problem. In this study, we explore the potential of ChatGPT as an information source for eliciting NFRs related to responsible fact-checking in journalism. Using a previously constructed reference model - developed by the first co-author during a series of Design Thinking sessions with journalism professionals - as a gold standard, three requirements engineering experts, none of whom were familiar with the domain, conducted individual chat sessions with ChatGPT and independently constructed Softgoal Interdependency Graphs (SIGs). Our findings go beyond a simple comparison with the gold standard. While some softgoals consistently emerged across sessions (e.g., trust, accuracy, transparency), participants also uncovered quality concerns such as integrity, dignity, and fairness - elements not explicitly included in the original model. These highlight risks that fact-checking practices must proactively mitigate and offer a broader understanding of relevant qualities in the domain. Additionally, the absence of certain softgoals from the LLM-generated models underscores the importance of human–AI collaboration to improve the completeness and contextual richness of SIGs.
The increasing complexity of software ecosystems (SECO) demands effective governance mechanisms to ensure long-term sustainability, particularly within proprietary SECO (PSECO). As organizations modernize their technology platforms, cloud migration emerges as a key strategy to improve scalability, flexibility, and operational resilience. However, selecting an appropriate migration approach for legacy software assets requires a structured decision-making process that aligns with governance principles, business continuity, and technical sustainability. This work investigates cloud migration strategies within a PSECO and proposes a decision tree artifact to assist IT managers in valuating migration options. We analyze real-world constraints, stakeholder concerns, and governance requirements through a participative case study in a large international organization. Our findings revealed that migration decisions are influenced not only by technical factors, such as architecture and performance, but also by business drivers, regulatory constraints, and organizational culture. The proposed guide provides a systematic approach to balancing modernization efforts with risk mitigation, helping organizations avoid technical debt while adapting to evolving industry demands. Using governance mechanisms in cloud migration strategies may support organizations in maintaining platform stability while fostering continuous innovation. Future research should explore how emerging cloud technologies and governance frameworks further impact modernization in the PSECO context.
La prueba de software es una fase crítica en el ciclo de desarrollo del software, ya que consiste en un conjunto de actividades diseñadas para evaluar la calidad de los productos que se construyen. Además, los datos obtenidos durante esta etapa fortalecen la confianza en el producto final y aportan información valiosa para respaldar la toma de decisiones por parte de todos los interesados. En la actualidad, el incremento de la complejidad de los productos que solicitan los clientes exige la implementación de estrategias innovadoras, orientadas a identificar la mayor cantidad de defectos y disminuir el riesgo de fallas. En este artículo, se propone un método para el proceso de diseño de casos de pruebas en formato Gherkin, tomando como entrada la Especificación de Requerimientos de Software y el Léxico Extendido del Lenguaje.
Los procesos de enseñanza y aprendizaje se caracterizan por reconocer el papel determinante del docente en identificar, planificar e instrumentar estrategias en la formación de sus estudiantes. La didáctica exige la utilización de estrategias y métodos centrados en el sujeto que aprende, en enfocar la enseñanza como un proceso de orientación del aprendizaje, donde los estudiantes no sólo se apropien de los conocimientos, sino que desarrollen diversas habilidades. La fase de Ingeniería de Requerimientos es crítica en el desarrollo de software, por lo que su enseñanza es fundamental. La formación académica debe generar habilidades y aptitudes determinadas como la abstracción, trabajo en equipo, toma de decisiones, comprensión del problema, etc., en donde se distinguen tres elementos: conceptual, procedimental e integrador. Para esto se necesitan técnicas que se incorporen en el proceso de enseñanza en el área de la ingeniería de requerimientos. La enseñanza de técnicas de ingeniería de requerimientos presenta desafíos como la dificultad de realizar prácticas reales. Metodología: para este caso de estudio planteamos la siguiente metodología: los estudiantes fueron divididos en grupos, algunos entrevistaron a un docente asistente de la cátedra mientras otros a un docente de otra carrera, quienes oficiaron de “clientes” a los efectos de obtener un documento de especificación de requerimientos. Resultados: el artículo presenta resultados parciales de nuevas estrategias didácticas para abordar la complejidad en la enseñanza de la ingeniería de requerimientos. La principal contribución radica en presentar nuevas estrategias que reflejen interacciones con los usuarios reales.
