Keywords

Requirements Engineering (RE), Artificial Intelligence and Natural Language Processing (NLP), Systems Modelling (UML, SysML).

Context

It's a well-known fact that software systems development is very often more expensive than expected, development time is difficult to evaluate and the end result disappoints stakeholders. One of the reasons for this is to be found in the upstream system design phase, known as Requirements Engineering (RE), which consists of gathering the expectations of project stakeholders from functional and non-functional points of view. Conventional methods associated to this stage are based on interviews, conducted by the project manager or a RE expert. They take the form of contractual documents in text format, which are difficult to verify and often ambiguous or incomplete. These discussions are therefore fundamental to building the future software system. In addition to interviews, rigorous approaches (goal models, prototyping, definition of user interfaces, use cases, context formalization) lead to set up requirements models as class or requirements diagrams [1], goal model [2], or social models [3]. Unfortunately, software engineering companies often neglect the RE stage, and don't take the necessary care with it.

Issues.
Collecting requirements focuses on understanding the context of the project to be developed and its objectives. It is above all an exercise based on natural language expression and comprehension. It has to be combined to a know-how about the required level of abstraction, so as not to get lost in details that would be more appropriate to the realization stage than the design stage. This know-how is also linked to two areas of expertise: expertise in RE (requirements collect, but also verification and validation) and knowledge about the target domain. In order to encapsulate the required expertise within a single entity, we wish to study the possibility of automating the collect of requirements and their validation/verification by a conversational agent (chatbot).

Solution to be explored.
Recent advances in AI techniques such as classification and NLP, based on Large Language Models, demonstrate the interest of such techniques for requirements elicitation and analysis [4 - 6]. Some work has been carried out on the improvement of requirements already formulated using user opinion classification techniques [7-9], and conversational agents are beginning to emerge [10]. As the results are far from satisfactory [11], and the use of AI techniques poses responsibility problems [12], we propose to study how to combine NLP techniques with the capitalization of a know-how relating to the RE and to the application domain (via an ontology, for example). This solution therefore requires experimentation with NLP and modelling techniques, and consequently a good knowledge of both fields. This work will be implemented and tested as part of a telehealth application design phase.

Objectives and expected results.
Several deliverables are expected from this thesis work:
− an analytical and critical study of the state of the art in both RE and NLP.
− an original method for improving not only the quality of requirements repositories and specification models, but also the process for developing them.
− a software prototype allowing the implementation of the method
− a proof of concept on a telehealth application
− publication of the work developed during the thesis an international conferences or international journals.

Applicant profile
Required diploma: Master degree in Computer Science.
Technical skills:
− Major Knowledge on Software Development Engineering and System Design
− Knowledge and practice in Model Driven Engineering: Meta Modelling, System Modelling and Model Transformation,
− Basic Knowledge in Machine Learning, − JAVA or Python programming, Soft skills:
− Solid oral and written communication capacity in French or in English
− Curiosity and Adaptability capacities to deal with the application domain: understanding knowledge and reasoning of experts in the telehealth domain.

Application
For scientific information or applicancy, contacts are:
gerard.dray@mines-ales.fr
anne-lise.courbis@mines-ales.fr
For administrative information, contact is: anne-catherine.denni@mines-ales.fr

Application must include:
− CV
− Master results (marks) per course
− Title and Supervisors of the Master internship
− Motivation letter pointing out your interest in the subject

Deadline for applicancy: May 31st 2024

Administrative Information
Institution: IMT Mines Alès (Ecole nationale supérieure des mines d’Alès), France
Laboratory: Euromov Digital Health in Motion, Univ Montpellier, IMT Mines Ales, France.
https://dhm.euromov.eu/
Doctoral school: I2S, Univ. Montpellier http://www.edi2s.univ-montp2.fr/
Speciality : Computer Science

Funding:
IMT Mines Alès contract
Contract duration: 36 months - Trial period: 2 months
Gross monthly salary: 2 272,50 euros
Working time: full time (151.67 hours / month)

Beginning: Sept. 2024 (or before Nov. 2024)

References.
[1] “OMG SYSML”, version 1.6. OMG Document, number formal/19-11-01, Nov. 2019.
[2] A. Van Lamsweerde, E. Letier. “From object orientation to goal orientation: A paradigm shift for requirements engineering,” International Workshop on Radical Innovations of Software and Systems Engineering in the Future. Berlin, Heidelberg: Springer Berlin Heidelberg, 2002.
[3] E.S. Yu. “Social modeling and I*,” Lect. Notes Comput. Sci, vol. 5600 LNCS, pp. 99–121, 2009.
[4] K. Liu, S. Reddivari, and K. Reddivari. “Artificial Intelligence in Software Requirements Engineering: State-of-the-Art,” Proc. - IEEE 23rd Int. Conf. Inf. Reuse Integr. Data Sci, pp. 106– 111, 2022.
[5] K. Papapanos, J. Pfeifer. “A literature review on the impact of artificial intelligence on in requirements Elicitation and Analysis,” Master report, University Stockholm, 2023.²
[6] S. Das, N. Deb, A. Cortesi, N. Chaki. “Extracting goal models from natural language requirement specifications,” in Journal of Systems and Software, 2024, vol. 211
[7] J. Wei, A. Courbis, T. Lambolais, B. Xu, P. L. Bernard, and G. Dray. “Zero-shot Bilingual App Reviews Mining with Large Language Models,” 35th IEEE International Conference on Tools with Artificial Intelligence (ICTAI), 2023.
[8] J. Wei, A. L. Courbis, T. Lambolais, P. L. Bernard, and G. Dray. “Towards boosting Requirements Engineering of a Health Monitoring App by Analysing Similar Apps: A Vision Paper, ” Proc. IEEE Int. Conf. Requir. Eng., pp. 75–80, 2022.
[9] P. Harth, O. Jähde, S. Schneider, N. Horn, and R. Buchkremer. “From Data to Human-Readable Requirements: Advancing Requirements Elicitation through Language-Transformer-Enhanced Opportunity Mining,” in Algorithms, vol. 16, no. 9, 2023.
[10] Walid Maalej. “From RSSE to BotSE: Potentials and Challenges Revisited after 15 Years,” IEEE/ACM 5th International Workshop on Bots in Software Engineering (BotSE), Melbourne, Australia, 2023, pp. 19-22.
[11] Z. Ji, N. Lee, R. Frieske, T. Yu, D. Su, Y. Xu, E. Ishii, Y. Ban, J. Ye, A. Madotto, P. Fung. “Survey of Hallucination in Natural Language Generation,” in ACM Comput. Surv., vol. 55, no. 12, 2023.
[12] W. Maalej, Y. D. Pham and L. Chazette, “Tailoring Requirements Engineering for Responsible AI,” in Computer, vol. 56, no. 4, pp. 18-27, April 2023.