Camille Salinesi

Abdelkarim El-Hajjami, Nicolas Fafin, Camille Salinesi

Recently, Large Language Models like ChatGPT have demonstrated remarkable proficiency in various Natural Language Processing tasks. Their application in Requirements Engineering, especially in requirements classification, has gained increasing interest. This paper reports an extensive empirical evaluation of two ChatGPT models, specifically gpt-3.5-turbo, and gpt-4 in both zero-shot and few-shot settings for requirements classification. The question arises as to how these models compare to traditional classification methods, specifically Support Vector Machine and Long Short-Term Memory. Based on five different datasets, our results show that there is no single best technique for all types of requirement classes. Interestingly, the few-shot setting has been found to be beneficial primarily in scenarios where zero-shot results are significantly low.

Abdelkarim El-Hajjami, Camille Salinesi

The shortage of publicly available, labeled requirements datasets remains a major barrier to advancing Artificial Intelligence for Requirements Engineering (AI4RE). While Large Language Models offer promising capabilities for synthetic data generation, systematic approaches to control and optimize the quality of generated requirements remain underexplored. This paper presents Synthline v1, an enhanced Product Line approach for generating synthetic requirements data that extends our earlier v0 version with advanced generation strategies and curation techniques. We investigate four research questions assessing how prompting strategies, automated prompt optimization, and post-generation curation affect data quality across four classification tasks: defect detection, functional vs. non-functional, quality vs. non-quality, and security vs. non-security. Our evaluation shows that multi-sample prompting significantly boosts both utility and diversity over single-sample generation, with F1-score gains from 6 to 44 points. The use of PACE (Prompt Actor-Critic Editing) for automated prompt optimization yields task-dependent results, greatly improving functional classification (+32.5 points) but reducing performance on others. Interestingly, similarity-based curation improves diversity but often harms classification performance, indicating that some redundancy may help ML models. Most importantly, our results show that synthetic requirements can match or outperform human-authored ones for specific tasks, with synthetic data surpassing human data for security (+7.8 points) and defect classification (+15.4 points). These findings offer practical insights for AI4RE and chart a viable path to mitigating dataset scarcity through systematic synthetic generation.

Abdelkarim El-Hajjami, Camille Salinesi

While modern Requirements Engineering (RE) heavily relies on natural language processing and Machine Learning (ML) techniques, their effectiveness is limited by the scarcity of high-quality datasets. This paper introduces Synthline, a Product Line (PL) approach that leverages Large Language Models to systematically generate synthetic RE data for classification-based use cases. Through an empirical evaluation conducted in the context of using ML for the identification of requirements specification defects, we investigated both the diversity of the generated data and its utility for training downstream models. Our analysis reveals that while synthetic datasets exhibit less diversity than real data, they are good enough to serve as viable training resources. Moreover, our evaluation shows that combining synthetic and real data leads to substantial performance improvements. Specifically, hybrid approaches achieve up to 85% improvement in precision and a 2x increase in recall compared to models trained exclusively on real data. These findings demonstrate the potential of PL-based synthetic data generation to address data scarcity in RE. We make both our implementation and generated datasets publicly available to support reproducibility and advancement in the field.

Nicolas Six, Nicolas Herbaut, Camille Salinesi

Designing blockchain-based applications is a tedious task. Compared to traditional software engineering, software architects cannot rely on previous experiences or proven practices, often formalized as software patterns. Also, the selection of an adequate blockchain technology is difficult without deep knowledge of the technology. This paper introduces Harmonica, a framework for the design and implementation of a blockchain-based application. This framework is divided in three parts: a decision-making engine to recommend a blockchain technology and blockchain-based software patterns relying on requirements, a configurator to generate code stubs and configuration files, and a knowledge base to support those tools.

Nicolas Six, Nicolas Herbaut, Camille Salinesi

A software pattern is a reusable solution to address a commonly occurring problem within a given context when designing software. Using patterns is a common practice for software architects to ensure software quality. Many pattern collections have been proposed for a large number of application domains. However, because of the technology's recentness, there are only a few available collections with a lack of extensive testing in industrial blockchain applications. It is also difficult for software architects to adequately apply blockchain patterns in their applications, as it requires deep knowledge of blockchain technology. Through a systematic literature review, this paper has identified 120 unique blockchain-related patterns and proposes a pattern taxonomy composed of multiple categories, built from the extracted pattern collection. The purpose of this collection is to map, classify, and describe all available patterns across the literature to help readers make adequate decisions regarding blockchain pattern selection. This study also shows potential applications of those patterns and identifies the relationships between blockchain patterns and other non-blockchain software patterns.

Nicolas Six, Claudia Negri Ribalta, Nicolas Herbaut et al.

Blockchain has been praised for its capacity to hold data in a decentralized and tamper-proof way. It also supports the execution of code through blockchain's smart contracts, adding automation of actions to the network with high trustability. However, as smart contracts are visible by anybody on the network, the business data and logic may be at risk, thus companies could be reluctant to use such technology. This paper aims to propose a pattern that allows the execution of automatable legal contract clauses, where its execution states are stored in an on-chain smart-contract and the logic needed to enforce it wraps it off-chain. An engine completes this pattern by running a business process that corresponds to the legal contract. We then propose a pattern-based solution based on a real-life use case: transportation of refrigerated goods. We argue that this pattern guarantees companies pseudonymity and data confidentiality while ensuring that an audit trail can be reconstituted through the blockchain smart-contract to identify misbehavior or errors. This paper paves the way for a future possible implementation of the solution described, as well as its evaluation.

Nicolas Six, Nicolas Herbaut, Camille Salinesi

Companies trying to build new solutions using blockchain are confronted with a plethora of available concurrent technologies that have many control knobs which require fine-tuning by experts. Exiting studies that build decision models for blockchain adoption or selection lack an automated way to use non-functional requirements to provide recommendations. In this paper, we build a knowledge base for blockchain solutions by analyzing whitepapers and studies, but also our benchmark results performed in a controlled environment. Then, we implement a Multi-Criterion Decision Analysis method to determine the most suitable blockchain solution from companies provided requirements and preferences. Finally, we illustrate our approach by running the decision process on a realistic supply-chain use case. This paper provides a rationale for blockchain deployment choices. While still limited in scope, we plan to include more blockchain alternative and more flexible requirements inputs in future work.

Asmaa Achtaich, Nissrine Souissi, Raul Mazo et al.

The Internet of Things (IoT) is the science of connecting multiple devices that coordinate to provide the service in question. IoT environments are complex, dynamic, rapidly changing and resource constrained. Therefore, proactively adapting devices to align with context fluctuations becomes a concern. To propose suitable configurations, it should be possible to sense information from devices, analyze the data and reconfigure them accordingly. Applied in the service of the environment, a fleet of devices can monitor environment indicators and control it in order to propose best fit solutions or prevent risks like over consumption of resources (e.g., water and energy). This paper describes our methodology in designing a framework for the monitoring and multi-instantiation of fleets of connected objects. First by identifying the particularities of the fleet, then by specifying connected object as a Dynamic Software Product Line (DSPL), capable of readjusting while running.

Christoph Becker, Ruzanna Chitchyan, Leticia Duboc et al.

Sustainability is a central concern for our society, and software systems increasingly play a central role in it. As designers of software technology, we cause change and are responsible for the effects of our design choices. We recognize that there is a rapidly increasing awareness of the fundamental need and desire for a more sustainable world, and there is a lot of genuine goodwill. However, this alone will be ineffective unless we come to understand and address our persistent misperceptions. The Karlskrona Manifesto for Sustainability Design aims to initiate a much needed conversation in and beyond the software community by highlighting such perceptions and proposing a set of fundamental principles for sustainability design.