REFSQ'24 Workshop, April 8th, 2024
Submit Here!
NLP4RE 2024 is over! We thank all participants, authors, and all others who contributed to the success of the workshop. Looking forward to meeting you all again at NLP4RE 2025!
Natural language processing (NLP) plays an essential role in several areas of software engineering (SE), and requirements engineering (RE) is no exception. Requirements are generally authored and communicated in textual form and different levels of formality, from structured (e.g., user stories) to unstructured natural language. In the last few years, the advent of massive and heterogeneous sources, such as tweets and app reviews, has attracted even more interest from the RE community, and the recent developments in large language models (LLMs) and generative AI have opened new opportunities for RE. LLMs will likely be the enabling technology for solving long-standing RE problems, such as traceability, classification, and compliance.
The main goal of NLP4RE is to represent a community- building venue for researchers who apply NLP technologies to solve RE problems and automatically support RE activities.
The NLP4RE Workshop is co-located with REFSQ'24. Check out the REFSQ'24 Conference here: https://2024.refsq.org
The workshop welcomes contributions regarding both theory and application of NLP technologies in RE. We also encourage contributions that highlight challenges faced by industrial practitioners when dealing with requirements expressed in NL, and the experiences of academics in technology transfer.
We are interested in Tool Papers (see the Call for Papers), in which the authors provide a brief description of an NLP tool for RE, and a plan for a tool demo at the workshop.
We are also interested in Report Papers (see the Call for Papers), in which the authors provide an overview on the current and past research of their teams. These contributions do not require novelty with respect to previous work, because the main goal of the workshop is to foster discussion and networking.
Moreover, this year we encourage submissions discussing the following topics:
Specific topics of interest include but are not limited to:
8 pages (plus 2 pages for references), describing novel technical solutions for the application of NLP technologies to RE-relevant problems in the topics of the workshop, or evaluation of existing solutions and NLP4RE-relevant phenomena using, e.g., experiments, and other empirically sound research strategies. This type of paper also includes case studies, interview/observational studies, and systematic literature reviews.
5 to 8 pages (plus 2 pages for references), describing practical experiences in the application of NLP technologies to RE-relevant artifacts. These papers should focus on retrospective reporting and discussion of lessons learned.
5 pages (plus 1 page for references), in which the authors provide an overview of their team's past, current or upcoming projects and research. These contributions include summaries of ongoing projects (EU, national, regional), which include an NLP4RE dimension, and preliminary project ideas for which the presenters are looking for collaborators. These contributions do not require novelty with respect to previous work, and are oriented to foster discussion and networking.
Our recommended template for Report Papers can be dowloaded here.
5 pages (plus 1 page for references), outlining a roadmap for research in the workshop’s topics, including industrial and research challenges based on currently available knowledge. We also encourage contributions highlighting challenges practitioners face when dealing with requirements expressed in NL, and challenges faced by academics in technology transfer studies or when applying/evaluating NLP4RE technologies in practice.
5 pages (including screenshots and references), in which the authors provide a short description of an NLP-based tool for RE with screenshots and a clear plan for a demo at the workshop. These contributions do not require novelty with respect to previous work, and the authors can also showcase tools presented in past conferences and workshops. These papers will be evaluated based on the potential interest raised by the tool, and based on the clarity of the plan for the demo.
1 page (including references), in which the authors present a paper previously presented at RE, REFSQ or ICSE. The contribution should briefly summarize the content of the original paper. These contributions are oriented to foster dissemination, and will not appear in the proceedings.
Submissions should be written in English and submitted in PDF format (page size A4, single column) formatted according to the CEUR Proceedings Style:
It is required for at least one author of each accepted submission to register, attend the workshop and present their research to the workshop participants.
All papers will undergo a traditional single-blind peer-review process (3 reviews per paper) to check scientific soundness, adequacy to the workshop topics, and compliance with the required template. We plan to publish the accepted papers in the CEUR Proceedings, with ISBN number.
Decoder-only LLMs, such as GPT, have revolutionized how we interact with artificial intelligence. Their ability to understand, generate, and manipulate language presents unprecedented opportunities and challenges across various disciplines, including Requirements Engineering (RE). This talk introduces a pioneering approach to integrating decoder-only LLMs into RE, poised to redefine the landscape of requirements elicitation, specification, and validation.
The presentation is structured into two primary segments. The first part delves into the application of decoder-only models in automating RE tasks. It explores how these models can assist in accurately capturing and specifying requirements, generating requirement documents, and automating the verification of requirements consistency and completeness. By examining case studies and current research, this section will highlight the transformative potential of decoder-only LLMs in enhancing efficiency, accuracy, and comprehensiveness in RE.
The second segment of the talk positions RE as a critical discipline for developing well-crafted prompts essential for interacting with decoder-only LLMs. It underscores the importance of precise, unambiguous, and comprehensive requirements in formulating prompts that elicit accurate and relevant responses from the models. This part will also discuss the art and science of crafting effective prompts, drawing parallels between requirements specification techniques and prompt engineering strategies.
Link to presentation slides is here.
Andreas Vogelsang is full professor for Software and Systems Engineering at the University of Cologne. He received a PhD from the Technical University of Munich. His research is at the intersection between requirements engineering, empirical software engineering, and machine learning. He has published over 100 papers in international journals and conferences such as TSE, JSS, IEEE Software, RE, and ICSE. In 2018, he was appointed as Junior-Fellow of the German Society for Informatics (GI). Further information can be obtained from https://cs.uni-koeln.de/sse.
9:00 - 9:05 - Introduction
9:05 - 10:05 - Keynote – Prompting the Future: Integrating Decoder-Only LLMs and Requirements Engineering by Andreas Vogelsang (slides)
10:05 - 10:35 - T-FREX: A Transformer-based Feature Extraction Method from Mobile App Reviews by Quim Motger, Alessio Miaschi, Felice Dell'Orletta, Xavier Franch and Jordi Marco.
11:00 - 11:30 - Requirements Classification for Smart Allocation: A Case Study in the Railway Industry by Sarmad Bashir, Muhammad Abbas and Mehrdad Saadatmand.
11:30 - 12:00 - Which AI Technique Is Better to Classify Requirements? An Experiment with SVM, LSTM, and ChatGPT by Abdelkarim El-Hajjami, Nicolas Fafin and Camille Salinesi.
12:00 - 12:30 - Automated Requirements Demarcation using Large Language Models: An Empirical Study by Kaishuo Wang, Feier Zhang and Mehrdad Sabetzadeh. (video)
14:00 - 14:30 - Documentation of non-functional requirements for systems with machine learning components by Elma Bajraktari, Thomas Krause and Christian Kücherer
14:30 - 15:00 - Peeking Outside the Black-Box: AI Explainability Requirements beyond Interpretability by Jakob Droste, Hannah Deters, Ronja Fuchs and Kurt Schneider
15:00 - 15:30 - RE4AI Panel by Workshop Participants
16:00 - 16:30 - Leveraging Knowledge Graphs for Goal Model Generation by Shahin Abdoul Soukour, William Aboucaya and Nikolaos Georgantas
16:30 - 17:00 - Automating Data Flow Diagram Generation from User Stories Using Large Language Models by Guntur Budi Herwanto
17:00 - 17:30 - Wrap-up and take away messages by Workshop Participants
For questions about the workshop, reach us via e-mail.
University of Luxembourg
(Luxembourg)
Monash
University
(Australia)
Utrecht University
(The Netherlands)
CNR-ISTI
(Italy)
University of Maine
(United States)
The continuity of the workshop is guaranteed by a steering committee, which currently consists of the following people: