Issue Tracking Systems (ITSs) serve multiple purposes in software development: they represent requirements and bugs, facilitate communication between team members, and support project management by assigning tasks. As software systems and projects increase in complexity, the number of work items and their relationships grows considerably, forming a complex network that is difficult and time-consuming for team members to navigate using traditional ITSs. Motivated by the interest of two companies in exploring the use of Artificial Intelligence in software development, we introduce the GraphRAG Dialogue Insights (GDI) framework. GDI uses knowledge graphs, Retrieval-Augmented Generation (RAG), and Large Language Models (LLMs) to enable users to query work items in natural language (NL). In addition to a NL response, GDI returns supporting information to help users understand how the LLMs queried the knowledge graph. We empirically validate GDI through two proof-of-concept implementations, each addressing a use case: i) onboarding and ii) trace link recovery. Initial user-centered validations indicate that practitioners find the system intuitive, useful, and accurate in generating responses. Moreover, they express willingness to adopt the system and identify additional potential use cases, including impact analysis, troubleshooting, and issue resolution. However, further refinement is required to deploy the system in practice.

Elicitation interviews are the most common requirements elicitation technique, and proficiency in conducting these interviews is crucial for requirements elicitation. Traditional training methods, typically limited to textbook learning, may not sufficiently address the practical complexities of interviewing techniques. Practical training with various interview scenarios is important for understanding how to apply theoretical knowledge in real-world contexts. However, there is a shortage of educational interview material, as creating interview scripts requires both technical expertise and creativity. To address this issue, we develop a specialized GPT agent for auto-generating interview scripts. The GPT agent is equipped with a dedicated knowledge base tailored to the guidelines and best practices of requirements elicitation interview procedures. We employ a prompt chaining approach to mitigate the output length constraint of GPT to be able to generate thorough and detailed interview scripts. This involves dividing the interview into sections and crafting distinct prompts for each, allowing for the generation of complete content for each section. The generated scripts are assessed through standard natural language generation evaluation metrics and an expert judgment study, confirming their applicability in requirements engineering training.

Abstract Interviewing stakeholders is the most popular technique for eliciting requirements. The success of an interview depends on the interviewer's theoretical knowledge, preparedness, and communication skills. Practice interviews allow students to apply their knowledge and improve their skills through experience. This practical training is resource‐intensive, requiring the time and effort of a stakeholder for each student, which may not be feasible for a large number of students. This paper introduces RoboREIT, an interactive Robotic tutor for Requirements Elicitation Interview Training. RoboREIT addresses the scalability problem of practice sessions with a robotic tutor acting as a stakeholder during the interview and providing feedback after the interview. We performed an exploratory user study to evaluate RoboREIT and demonstrate its applicability in requirements elicitation interview training. The quantitative and qualitative analyses of the users' responses reveal the appreciation of RoboREIT. Our study is the first in the literature that utilizes a social robot in requirements elicitation interview education. RoboREIT's design incorporates replaying faulty interview stages and allows the student to learn from mistakes by a second time practicing. All participants praised the feedback component, which is not present in the state of the art, for being helpful in identifying the mistakes. A favorable response rate of 81% for the system's usefulness indicates the positive perception of the participants.

Natural language processing (NLP) techniques have been widely applied in the requirements engineering (RE) field to support tasks such as classification and ambiguity detection. Despite its empirical vocation, RE research has given limited attention to replication of NLP for RE studies. Replication is hampered by several factors, including the context specificity of the studies, the heterogeneity of the tasks involving NLP, the tasks’ inherent hairiness , and, in turn, the heterogeneous reporting structure. To address these issues, we propose a new artifact, referred to as ID-Card , whose goal is to provide a structured summary of research papers emphasizing replication-relevant information. We construct the ID-Card through a structured, iterative process based on design science. In this article: (i) we report on hands-on experiences of replication; (ii) we review the state-of-the-art and extract replication-relevant information: (iii) we identify, through focus groups, challenges across two typical dimensions of replication: data annotation and tool reconstruction; and (iv) we present the concept and structure of the ID-Card to mitigate the identified challenges. This study aims to create awareness of replication in NLP for RE. We propose an ID-Card that is intended to foster study replication but can also be used in other contexts, e.g., for educational purposes.

Abstract Context Automated classifiers, often based on machine learning (ML), are increasingly used in software engineering (SE) for labelling previously unseen SE data. Researchers have proposed automated classifiers that predict if a code chunk is a clone, if a requirement is functional or non-functional, if the outcome of a test case is non-deterministic, etc. Objective The lack of guidelines for applying and reporting classification techniques for SE research leads to studies in which important research steps may be skipped, key findings might not be identified and shared, and the readers may find reported results (e.g., precision or recall above 90%) that are not a credible representation of the performance in operational contexts. The goal of this paper is to advance ML4SE research by proposing rigorous ways of conducting and reporting research. Results We introduce the ECSER (Evaluating Classifiers in Software Engineering Research) pipeline, which includes a series of steps for conducting and evaluating automated classification research in SE. Then, we conduct two replication studies where we apply ECSER to recent research in requirements engineering and in software testing. Conclusions In addition to demonstrating the applicability of the pipeline, the replication studies demonstrate ECSER’s usefulness: not only do we confirm and strengthen some findings identified by the original authors, but we also discover additional ones. Some of these findings contradict the original ones.

User stories are widely used to capture the desires of the users in agile development. A set of user stories is easy to read and write but incapable of representing the hierarchical relations and synergies among the user stories. By contrast, goal models are uncommon in industrial projects however they can express the structure and other relations among requirements captured as goals. This paper presents ArTu, a tool for generating goal models from user stories to effortlessly benefit from both. Given a set of user stories, our tool generates goal models with different structures depending on the heuristic selected by the user. Users can import, edit, and export model data in different formats.

User stories are commonly used to capture user needs in agile methods due to their ease of learning and understanding. Yet, the simple structure of user stories prevents us from capturing relations among them. Such relations help the developers to better understand and structure the backlog items derived from the user stories. One solution to this problem is to build goal models that provide explicit relations among goals but require time and effort to build. This paper presents a pipeline to automatically generate a goal model from a set of user stories by applying natural language processing (NLP) techniques and our initial heuristics to build realistic goal models. We first parse and identify the dependencies in the user stories, and store the results in a graph database to maintain the relations among the roles, actions, and objects mentioned in the set of user stories. By applying NLP techniques and several heuristics, we generate goal models that resemble human-built models. Automatically generating models significantly decreases the time spent on this tedious task. Our research agenda includes calculating the similarity between the automatically generated models and the expert-built models. Our overarching research goals are to provide i. an NLP-powered framework that generates goal models from a set of user stories, ii. several heuristics to generate goal models that resemble human-built models, and iii. a repository that includes sets of user stories, with corresponding human-built and automatically generated goal models.

Requirements classification is a traditional application of machine learning (ML) to RE that helps handle large requirements datasets. A prime example of an RE classification problem is the distinction between functional and non-functional (quality) requirements. State-of-the-art classifiers build their effectiveness on a large set of word features like text n-grams or POS n-grams, which do not fully capture the essence of a requirement. As a result, it is arduous for human analysts to interpret the classification results by exploring the classifier's inner workings. We propose the use of more general linguistic features, such as dependency types, for the construction of interpretable ML classifiers for RE. Through a feature engineering effort, in which we are assisted by modern introspection tools that reveal the hidden inner workings of ML classifiers, we derive a set of 17 linguistic features. While classifiers that use our proposed features fit the training set slightly worse than those that use high-dimensional feature sets, our approach performs generally better on validation datasets and it is more interpretable.

Over the last two decades, much attention has been paid to the area of goal-oriented requirements engineering (GORE), where goals are used as a useful conceptualization to elicit, model, and analyze requirements, capturing alternatives and conflicts. Goal modeling has been adapted and applied to many sub-topics within requirements engineering (RE) and beyond, such as agent orientation, aspect orientation, business intelligence, model-driven development, and security. Despite extensive efforts in this field, the RE community lacks a recent, general systematic literature review of the area. In this work, we present a systematic mapping study, covering the 246 top-cited GORE-related conference and journal papers, according to Scopus. Our literature map addresses several research questions: we classify the types of papers (e.g., proposals, formalizations, meta-studies), look at the presence of evaluation, the topics covered (e.g., security, agents, scenarios), frameworks used, venues, citations, author networks, and overall publication numbers. For most questions, we evaluate trends over time. Our findings show a proliferation of papers with new ideas and few citations, with a small number of authors and papers dominating citations; however, there is a slight rise in papers which build upon past work (implementations, integrations, and extensions). We see a rise in papers concerning adaptation/variability/evolution and a slight rise in case studies. Overall, interest in GORE has increased. We use our analysis results to make recommendations concerning future GORE research and make our data publicly available.

Context. Goal models have long been critiqued for the time it takes to construct them as well as for their limited cognitive and visual scalability. Is such criticism general or does it depend on the supported task? Objectives. We advocate for the latter and the aim of this paper is to demonstrate that the next release problem is a suitable application domain for goal models. This hypothesis stems from the fact that product release management is a long-term investment, and software products are commonly managed in "themes" which are smaller focus areas of the product. Methods. We employ a version of goal models that is tailored for the next release problem by capturing requirements, synergies among them, constraints, and release objectives. Such goal model allows discovering optimal solutions considering multiple criteria for the next release. Results. A retrospective case study confirms that goal models are easier to read and comprehend when organized in themes, and that the reasoning results help product managers decide for the next release. Our scalability experiments show that, through reasoning based on optimization modulo theories, the discovery of the optimal solution is fast and scales sufficiently well with respect to the model size, connectivity, and number of alternative solutions.

Dealing with multiple requirement failures is an essential capability for self-adaptive software systems. This capability becomes more challenging in the presence of conflicting goals. This paper is concerned with the next adaptation problem: the problem of finding the best next adaptation in the presence of multiple failures. ‘Best’ here means that the adaptation chosen optimizes a given set of objective functions, such as the cost of adaptation or the degree of failure for system requirements. The paper proposes a formal framework for defining the next adaptation problem, assuming that we can specify quantitatively the constraints that hold between indicators that measure the degree of failure of each requirement and control parameters. These constraints, along with one or several objective functions, are translated into a constrained multi-objective optimization problem that can be solved by using an OMT/SMT (Optimization Modulo Theories/Satisfiability Modulo Theories) solver, such as OptiMathSAT. The proposed framework is illustrated with the Meeting Scheduler exemplar and a second, e-shop case study.

Risks of software projects are often ignored and risk analysis is left for later stages of project life-cycle, resulting in serious financial losses. This paper proposes a goal-oriented risk analysis framework that includes inter-dependencies among treatments and risks in terms of likelihood and generate optimal solutions with respect to multiple objectives such as goal rewards, treatment costs, or risk factor. The Loan Origination Process illustrates our approach and a detailed analysis of the visual notation is provided.

Over the last two decades, much attention has been paid to the area of Goal-Oriented Requirements Engineering(GORE), where goals are used as a useful conceptualization to elicit, model and analyze requirements, capturing alternatives and conflicts. Goal modeling has been adapted and applied to many sub-topics within RE and beyond, such as agent-orientation, aspect-orientation, business intelligence, model-driven development, security, and so on. Despite extensive efforts in this field, the RE community lacks a recent, general systematic literature review of the area. As a first step towards providing a GORE overview, we present a Systematic Literature Map, focusing on GORE-related publications at a high-level, categorizing and analyzing paper information in order to answer several research questions, while omitting a detailed analysis of individual paper quality. Our Literature Map covers the 246 top-cited GORE-related conference and journal papers, according to Scopus, classifying them into a number of descriptive paper types and topics, providing an analysis of the data, which is made publicly available. We use our analysis results to make recommendations concerning future GORE research.

Sociotechnial systems (STSs) consist of a complex interplay of technical components, humans, and organizations. As other types of systems, STSs need to evolve in response to changing requirements and operational environments. Evolving STSs is a complex activity, which requires reconfiguration of technical components as well as rerouting of interactions among human and social actors. Moreover, reconfiguration has to respect participant autonomy, while coping with conflicting goals and noncooperation in identifying a configuration that minimizes changes relative to the current configuration. In this paper, we present a framework that supports design and evolution of STSs. The framework includes (i) the DEST language for modeling STSs as goal-oriented actors that interact via social commitments; (ii) techniques for building a network of interactions that fulfills participant requirements; and (iii) techniques for evolving an existing STS while minimizing change. We encode the design and evolution of STSs as an automated planning problem.

We address the challenge of requirements engineering for sociotechnical systems, wherein humans and organizations supported by technical artifacts such as software interact with one another. Traditional requirements models emphasize the goals of the stakeholders above their interactions. However, the participants in a sociotechnical system may not adopt the goals of the stakeholders involved in its specification. We motivate, Protos, a requirements engineering approach that gives prominence to the interactions of autonomous parties and specifies a sociotechnical system in terms of its participants' social relationships, specifically, commitments. The participants can adopt any goal they like, a key basis for innovative behavior, as long as they interact according to the commitments. Protos describes an abstract requirements engineering process as a series of refinements that seek to satisfy stakeholder requirements by incrementally expanding a specification set and an assumption set, and reducing requirements until all requirements are accommodated. We demonstrate this process via the London Ambulance System described in the literature.