Leveraging Retrieval-augmented LLMs for Automated Test Case Generation from Software Requirements Specification

The GOHIT platform is a web-based system designed to support the management of student competitions within higher-education institutions. The first version of the platform lacked structured requirements documentation, resulting in unclear user roles and inconsistent workflows. This study aimed to establish a lightweight yet disciplined requirements-engineering (RE) process for the redevelopment of GOHIT v2. Using a qualitative case-study approach, data were collected through semi-structured interviews and document analysis involving three key stakeholders: the founder, the developer, and the program advisor. Thematic analysis was applied to transform stakeholder inputs into structured, verifiable requirements based on IEEE 830 documentation principles. The process yielded seventeen functional requirements, each traceable to stakeholder sources and internally evaluated using IEEE 830 quality attributes. The resulting Software Requirements Specification demonstrated high levels of completeness, consistency, and traceability, confirming that a structured yet adaptable RE approach can be effective for small or academic development teams. This study contributes a replicable model for implementing lightweight requirements engineering in resource-limited settings and establishes a baseline for future validation and quality assurance using a test-based mechanism.

The development of software applications involves translating software requirement specifications (SRS) into structured models that guide system design. Among these, sequence diagrams are essential for visualizing dynamic interactions, but their manual construction from natural language descriptions is often error-prone and time-consuming. This study proposes an automated method for extracting sequence diagram elements namely classes, subclasses, and attributes from scenario sections of SRS documents. The approach leverages Natural Language Processing (NLP) techniques, combining Bidirectional Encoder Representations from Transformers (BERT) for contextual embeddings and Support Vector Machine (SVM) for classification. Noun phrases are identified and classified into UML-relevant entities using this hybrid model. To evaluate performance, two datasets SIData and SILo were used, each exhibiting distinct textual styles and domain characteristics. The system’s effectiveness was assessed using standard evaluation metrics such as precision, recall, and F1-score. Results indicate that the method is capable of capturing contextual relationships between extracted elements, although its performance varies across datasets, suggesting the need for further refinement. Overall, the study contributes toward automating early software design phases and reducing manual modeling effort.

Accurate classification of software requirements is a crucial task in Software Engineering (SE) that prioritizes development efforts and ensures the holistic quality of the system, encompassing both Functional Requirements (FRs) and Non-Functional Requirements (NFRs). While the majority of requirements classification research has so far focused on binary classification, fine-grained multi-class classification still encounters the challenge of extreme class imbalance in requirements datasets. To mitigate this limitation, we present TextAttack Oversampling (TAOS), a novel method that utilizes a Natural Language Processing (NLP)-based text augmentation technique to address this imbalance, thus reducing dependence on expensive expert labeling. An empirical assessment of a 12-class requirements dataset indicates that TAOS considerably outperforms standard classification techniques. Our approach achieves a 24% gain in F1-score, increasing from 0.75 to 0.93, and effectively improves the performance on minority requirement classes that were previously undetectable. This study demonstrates the effectiveness of context-aware text augmentation in improving multi-class requirements classification, thus providing a proven method to enhance the reliability and usefulness of automated requirements analysis and management tools for software engineers.

Software requirement specifications (SRS) often contain repetitive, ambiguous, or inconsistent requirements, which drives up costs and delays project timelines. Because they mainly rely on syntactic similarity, traditional redundancy detection methods like TF-IDF and Word2Vec have trouble detecting semantic overlaps. This study proposes a semantic pruning framework that uses advanced NLP techniques, with a focus on transformer-based models like BERT, to find and eliminate superfluous requirements from SRS documents. Precision, recall, F1-score, and runtime were used as evaluation criteria to compare several methods, including CountVectorizer, TF-IDF, Word2Vec, and BERT. The findings show that deep learning models outperform conventional methods, which yield high precision but poor recall. Despite having a longer runtime, BERT outperformed Word2Vec with F1 = 0.87 and recall = 0.77. The outcomes show the effectiveness of transformer-based embeddings for re-dundancy detection and provide a scalable approach to improve SRS quality while reducing the amount of manual review effort

This research focuses on the development of a freelance e-commerce website for WardellTech using the Laravel framework. The objective is to build a web-based platform that is secure, responsive, and user-friendly, supporting freelance service transactions. Laravel was selected for its structured MVC architecture, built-in security features, and its ability to streamline development and maintenance. A Software Requirements Specification (SRS) document was created to define both functional and non-functional requirements in detail, ensuring that all stakeholders—developers, testers, and project managers—have a unified understanding of the system. WardellTech specializes in web and mobile application development, emphasizing quality design, user experience, and data security. Therefore, the website integrates key features such as secure payment gateways, intuitive navigation, and optimized performance. Continuous analysis and improvement of the user interface are also prioritized to enhance usability and customer satisfaction. The final system is expected to not only meet WardellTech’s business needs but also compete effectively in the freelance digital services market. Additionally, it is designed to adapt to ongoing technological advancements and evolving user demands in the dynamic e-commerce landscape.

This paper presents an ontology developed to address challenges such as com munication gaps, risks of errors, and inconsistencies during the manual process of creating software requirement specifications (SRS). The proposed ontology offers a systematic and formal depiction of the requirements, enhancing consis tency and communication among stakeholders. The ontology has been devel oped from the software requirements documents to facilitate the development process. This paper discusses the process of creating the ontology and demon strates using Pellet Reasoner for inference and Prot´eg´e for ontology construction to save and reuse information. The ontology seems to be efficient in manag ing complex software projects, enabling accurate requirement retrieval through SPARQL queries. This study emphasizes how incorporating ontologies into re quirement engineering can significantly enhance the quality and reliability of SRS.

This research presents a design of software requirements specifications for Bengkel Sat-Set, an Android-based mobile application designed to simplify the digital ordering and management of vehicle service. This application is designed as a solution to various problems often encountered in conventional repair shop systems, such as long queues, lack of transparency regarding repair status, and limited access to vehicle service history. With this application, users are expected to be able to order services online, monitor vehicle repair status in real-time, and view service history without having to visit the repair shop in person. This system design uses a software engineering approach, with technical specifications using the Kotlin programming language for developing the Android interface and Firebase as the backend service. Firebase was chosen for its ability to provide efficient and secure data storage, user authentication, and real-time data communication. This research focuses on the early stages of development, namely designing the system's functional and non-functional requirements, including user flows and initial architecture. The primary objective of this paper is to provide a conceptual and technical foundation as part of the digital transformation process in the automotive service sector. It is hoped that through the design of this Sat-Set Workshop application, workshop service providers can improve operational efficiency, expand service access, and increase customer satisfaction through the use of adaptive and user-friendly digital technology.

In case report studies (adverse effects of drugs/vaccines) that are unstructured, a structured relational model is not applicable for designing a database, necessitating the use of an unstructured data model, specifically NoSQL. Therefore, the most important and optimal unstructured data model for eCRF, which has the nature of a form, is the document-oriented model. This paper develops and evaluates a reporting system for drug intervention studies with high variability in adverse events that utilizes a document-based NoSQL data model and the eCRF nature, allowing for the management of structured and unstructured data. The main objective of the research is to create a flexible, fast, and efficient system for collecting and analyzing data related to drug/vaccine adverse events, especially the COVID-19 vaccine, for stakeholders. This research is of an applied-descriptive type. In this research, first, after studying library resources, the requirements and requirements for the design of the proposed system were determined in the form of the Software requirements specification (SRS) standard. Then, the design and implementation included modeling and creating a prototype of the web-based system. To evaluate usability, the User Experience Questionnaire (UEQ) was used, system security was assessed using the Application Security Verification Standard (ASVS) questionnaire, and a comparative evaluation of the performance between MongoDB and SQLServer was performed. This research was conducted with the aim of designing and evaluating a reporting system for COVID-19 vaccine side effects, based on a document-oriented data model. It includes key components such as the information and side effects collection module, the document management module, and the reporting module. The results indicated that the user experience of the system, in terms of attractiveness, transparency, efficiency, reliability, motivation, and innovation, had an average score of 2.31, placing it within the top 10% of results. Additionally, the evaluations showed that the system employs effective security controls; however, improvements were needed in certain areas such as meeting management and authentication. A comparative assessment of the performance between the document-oriented data model and the relational data model demonstrated that the proposed system was able to provide better performance in response time and management of unstructured data. Evaluations have shown that utilizing case report forms, along with the advantages of a document-oriented data model, can be effective in collecting the minimum necessary data set for interventional studies, particularly those related to drug side effects such as the COVID-19 vaccine. Given the variable nature of the virus and the potential for unknown side effects, this requires flexible and precise approaches. Additionally, the use of the existing system, considering the results of the security and usability assessment, could be effective if access to external systems is improved.

Tranformasi dalam layanan kesehatan mendorong pengembangan aplikasi layanan rumah sakit sebagai solusi inovatif untuk meningkatkan aksesibilitas layanan rumah sakit. Penelitian ini menganalisis bagaimana penerapan metode waterfall hybrid dalam pengembangan aplikasi layanan rumah sakit yang mengalami perubahan signifikan dari awalnya berbasis web menjadi pendekatan mobile first. Dengan menggunakan pendekatan kualitatif melalui studi dokumen dan wawancara mendalam, penelitian ini mengevaluasi bagaimana metode hybrid dalam mengintegrasikan metode waterfall yang terstruktur dalam pengembangan fitur inti dan iterasi agile untuk pengembangan fitur dinamis. Hasil penelitian ini menunjukkan bahwa 60-63% fitur inti (manajemen data dan master data) berhasil di pertahankan sesuai dengan spesifikasi awal atau Software Requirements Specification, sementara 25-40% fitur antarmuka pengguna berhasil diadaptasi ke mobile, dan 9-12,5% fitur baru ditambahkan untuk merespons permintaan dari stakeholder. Tantangan utama meliputi modularisasi dokumen dan kordinasi tim, yang diadaptasi dengan strategi living document, integrasi rapat mingguan, penggunaan alat manajemen ganda (gantt chart dan scrum board). Penelititan ini menunjukkan bahwa penggunaan metode waterfall hybrid efektif untuk proyek dengan perubahan dinamis namun tetap memerlukan kontrol ketat, memberikan kerangka terhadap pengembang aplikasi kesehatan digital yang adaptif tetapi tetap menjaga stabilitas dan kepatuhan regulasi.

For decades, it has been proven that software testing is a vital component of the software development lifecycle and ensures reliability, functionality, and performance. However, traditional test case generation methods face challenges such as high time and resource demands and susceptibility to human error, especially in large-scale and complex software systems.The paper provides an extensive exploration of artificial intelligence (AI) applications in software test case generation, focusing on analyzing current industry practices and creating a predictive model designed to optimize this critical aspect of software quality assurance. To address these limitations, the adoption of AI techniques for automating and improving test case generation has gained significant traction. This research pursues two key objectives: first, to thoroughly analyze existing AI-driven testing techniques and strategies for test case generation through an extensive review of academic literature, industry reports, and case studies. This analysis delves into search-based, machine-learning approaches and natural language processing (NLP) techniques. Furthermore, it evaluates their application across different testing levels—unit, integration, system, and acceptance testing—and software domains like web applications, mobile platforms, embedded systems, and safety-critical environments. The analysis highlights current industry practices and identifies areas where AI can significantly enhance efficiency and effectiveness in software testing.The second objective involves designing and implementing a predictive model for optimal test case generation using advanced AI techniques. The model employs machine learning frameworks, including deep learning architectures like recurrent neural networks (RNNs), long short-term memory (LSTM) networks, and transformer-based models. By training on diverse datasets, including historical test data, software requirements specifications (SRS), source code, and execution logs, the model ensures broad applicability. Its focus includes maximizing test coverage, minimizing test suite size, and prioritizing test cases based on their fault-revealing potential, making the testing process more efficient and effective. The architecture accommodates various input formats, enabling a comprehensive, context-aware test case generation process.This research makes a significant contribution to software testing by offering a detailed analysis of AI-driven test case generation practices and introducing a robust predictive model to address existing challenges. The findings present practical solutions for software development professionals and researchers, improving software quality, reducing costs, and accelerating development timelines.

The MBKM (Merdeka Belajar Kampus Merdeka) policy is motivated to prepare students to know social, cultural, world of work and technological advances that have been so rapid. Competence of students must be prepared to answer the needs of the times. Link and match activities are one of the strategies used, not only with the business world and the industrial world, but also with a future that is changing so rapidly. Information Systems Study Program has implemented several programs contained in the guidelines for the implementation of the Independent Learning Campus, although on a limited scale, namely the cross-study student exchange program and the Internship program. It is necessary to make adjustments to the programs that have been implemented in Study Program to adjust the curriculum of the study program and to survey the academic community. This survey needs to be carried out to find out how far the responses, responses and knowledge of the academic community towards MBKM activities. From the results of the survey, the Study Program followed up by designing an information system for the development and strengthening of the results of the academic community for MBKM activities in the form of a Software Requirement Specification. With this Information System Design, it is hoped that it will provide convenience for the development of the Integrated MBKM information system so that it can provide convenience for the academic community in carrying out MBKM activities within the Study Program environment and also to support MBKM governance and management.

ABSTRACTNatural language processing (NLP) offers the potential to automate quality assurance of software requirement specifications. In particular, large‐scale projects involving numerous suppliers can benefit from this improvement. However, due to privacy restrictions especially in highly restrictive industries, the availability of software requirements specification documents for training NLP tools is severely limited. Also, domain‐ and project‐specific vocabulary, as such in the aerospace domain, require specialized models for processing effectively. To provide a sufficient amount of data to train such models, we studied algorithms for the augmentation of textual data. Four algorithms have been investigated by expanding a given set of requirements from the European Space projects generating correct and incorrect requirements. The initial study yielded data of poor quality due to the particularities of the domain‐specific vocabulary, yet laid the foundation for the algorithms' improvement, which, eventually, resulted in an increased set of requirements, which is 20 times the size of the seed set. A complementing experiment demonstrated the usability of augmented requirements to support AI‐based quality assurance of software requirements. Furthermore, a selected improvement of the augmentation algorithms demonstrated notable quality improvements by doubling the number of correctly augmented requirements.

The rising frequency of flash floods due to climate change demands efficient detection systems to reduce their impact. This study presents the "Early Flash Flood Detection System Using Machine Learning Algorithms," which integrates IoT sensors and machine learning for accurate, real-time flood prediction. Developed with Agile methodology, the project utilized key technologies like Flutter and TensorFlow to enhance functionality and user engagement. Testing showed 72% prediction accuracy, demonstrating the system's potential as a scalable solution for disaster management, advancing public safety, and fostering resilient communities.

Abstract This article presents a detailed analysis of the Software Requirement Specification (SRS) and Use Case Diagram for the Metaverse Museum Muhammadiyah, a virtual museum platform designed to offer an immersive and interactive user experience. The study aims to assess the alignment between the functional requirements defined in the SRS and their representation in the Use Case Diagram, ensuring that all critical functionalities are accurately captured and adequately addressed in the system design. The analysis identifies key functional requirements such as user navigation, interaction, content management, and customization features, evaluating their presence and clarity within the Use Case Diagram. Additionally, the study examines the non-functional requirements, including availability, portability, response time, safety, security, and ergonomic design, highlighting their importance in delivering a reliable and user-friendly system. The findings reveal a strong correlation between the SRS and the Use Case Diagram, with most functionalities well-represented. However, the analysis also uncovers certain gaps such as the omission of features like link sharing which require further refinement in the diagram. The article concludes by emphasizing the need for continuous improvement in both the SRS and Use Case Diagram to ensure comprehensive coverage of user needs and to meet the high standards expected in a metaverse environment. This analysis serves as a critical step in guiding the successful development and implementation of the Metaverse Museum Muhammadiyah.

Few-Shot Evaluation of Vision Language Models for Detecting Visual Defects in Autonomous Vehicle Software Requirement Specifications

In the digital age, students face increasing demands to manage various academic assignments and deadlines effectively. Manual task management often leads to missed deadlines, irregularities, and reduced productivity. Therefore, a structured, easy-to-use task monitoring system is needed to support learning efficiency. This research explores Task Minder, a task monitoring system that helps individuals and organizations manage tasks in a structured manner. With user-friendly features such as account creation, task editing, status tracking, and automatic reminders, Task Minder aims to increase productivity and efficiency. A descriptive method is used to describe and analyze this system, through a whitebox approach in the creation of Software Requirements Specification (SRS) documents. Data was collected through interviews, observations, and documentation, then analyzed qualitatively using thematic analysis methods. The results show an increase in user productivity, but also identify some limitations that need to be considered for effective use. Overall, Task Minder makes a real contribution as a digital solution in academic task management, helping students organize work more efficiently, responsively, and integrated with learning needs.

Disagreements in interpreting words in software requirements specifications (SRSs) can lead to project failure. Various approaches to identifying and preventing ambiguous words in SRSs have been proposed. Yet, it is unclear which ambiguous words are used in the actual SRSs and to what extent they need to be modified. This paper quantitatively analyzes existing SRSs to clarify (1) how many ambiguous words are included in SRSs and (2) how many of these words require correction. This paper targets the Request for Proposals (RFPs), which describe the initial requirements of 40 systems of local governments, libraries, universities, and hospitals in Japan. Ten ambiguous Japanese words were analyzed. The result shows that “juubun” (sufficient) appeared most frequently, and 42% required correction when this word was used. The result also indicates that the number of ambiguous words varied greatly among the RFPs and that larger RFPs did not necessarily contain more ambiguous words.

Software Requirements Review (SRR) is a formal review process in which several reviewers read all or parts of the Software Requirements Specification (SRS) to look for defects in the requirements. During requirements review sessions, reviewers may employ various reading techniques to ensure that the requirements have been completely and clearly specified. To review an SRS document, the review leader must organise a review session, the reviewers must physically meet and provide their review feedback during the session. In these situations, the review leader must schedule the review session based on the reviewers' availability, which can be laborious and time-consuming to arrange. Additionally, the review leader needs to manually consolidate all the outcomes of the review session, which could also require a considerable amount of effort and time. However, there has been insufficient research to identify the effectiveness of reading techniques for requirement reviews by employing a dedicated tool support for requirements review. Using a web-based application, called e-Review, the aim of this study is to experiment the effectiveness of Checklist-based Reading (CBR) and Perspective-based Reading (PBR) techniques during requirements review session.

Software Requirements Specifications is a document that describes the requirements that occur in the development of a software system. The category of requirements is defined in two types: Functional Requirements (FR) and Non-Functional Requirements (NFR). Software Requirements Engineering is critical in successfully designing a piece of software. Many studies have examined the classification of software requirements using machine learning, but none have compared bagging algorithms with Support Vector Machine (SVM). This study compares text feature extraction techniques with machine learning algorithms Bagging and Support Vector Machine to solve the Software Requirement Classification problem. Using vectorization techniques from word2vec: Continuous Bag of Words and Skip-gram can help produce the best model performance for Bagging and SVM models. In this study, the data used is expansion data from the PROMISE repository, namely PROMISE_exp, the repository is a collection of software requirements data that has been labeled. To measure performance, this study uses an evaluation matrix, namely precision, recall and f1-score. As a result, the two models that have been trained using the Continuous Bag of Words and skip-gram vectorization techniques will be compared to determine the more optimal model for classifying software requirements from the promise_exp repository.