Keywords: behaviour-driven development, model-based testing, behaviour-driven testing
Astract: Behaviour-driven development (BDD) is a strategy to describe a system’s specification using a business domain language: “Given a precondition; “When” some action is performed; “Then” an outcome is achieved'' (GWT). This style improves the communication between the involved parties and the resulting specification can be leveraged to drive automated tests. While this approach works well in practice, it suffers from some disadvantages: It is informal and verbose, thus bearing the risk of failing to specify all parts of a system.Model-based testing (MBT) is a structured approach for automatically generating test cases. It is well-suited for describing complex interactions and cross-linked code paths using models. Even basic graphical state machines can define what takes many pages to write down in natural language. Models describe a system on a higher level of abstraction and allow to quickly recreate test cases in the event that the behaviour of the system changes or gets extended.This thesis presents a testing approach that combines BDD with MBT based on state machine models to automate the process of writing BDD tests. A prototype generating BDD tests from graphical state machine models has been developed and evaluated in a case study with promising results: The generated test cases covered the functionality of the tested system, and the effort to create them was comparable to writing similar test cases by hand. A survey among practitioners showed that while users were able to identify automatically generated BDD tests, in some instances, they preferred them over manually written ones.

Keywords:
Astract: This year´s 39th ER conference is dedicated to a topic that represents a phenomenonunprecedented in the history of humankind. The digital transformation encompasses allareas of life and work. It is accompanied by new types of services, new forms ofdivision of labor, interpersonal interaction, and international cooperation. It thus has adirect impact on how we see the world and what perspectives we develop for our futurelives. Last but not least, we can assume that the ongoing digitalization will also have alasting impact on scientific research. Conceptual modeling is of central importance forthe successful management of the digital transformation. On the one hand, all areas oflife and work are increasingly permeated by software. Conceptual models are requirednot only for the development of software, but also for the appropriate structuring ofdata. They promote reuse, integration, and integrity. Furthermore, conceptual modelsare also suitable for supporting the use of software. They help to open the black box asto which software often presents itself and thus contribute to transparency and userempowerment. At the same time, the digital transformation also brings with it specificchallenges for modeling research. In order to support the design of software that can beadapted to profound changes of requirements, powerful abstractions are needed that arebeyond the capabilities of today´s prevalent modeling languages. In addition, AIresearch, especially in thefield of machine learning, is associated with aquasi-existential challenge of modeling research. Thus, some proponents of AI researchalready foresee the end of traditional conceptual modeling. It would last too long andwould be too expensive. It could be better handled by machines. Such daringhypotheses may be seen as a threat. But above all they are an occasion to reflect onfundamental questions of conceptual modeling, such as the difference between con-cepts and classifications or between human thought and data processing. Probably thecentral question is not whether and when machine learning can take over the humanactivity of conceptual modeling, but how the inductive analysis of large amounts ofdata and human abstraction can be synergistically combined.Given the fascination that the digital transformation holds for conceptual modelingresearch, it is not surprising that we were able to quickly agree on this conference topicduring last year´s ER conference in Salvador, Brazil. At that time, none of us had anyidea that the digital transformation would be significant for the conference in a com-pletely different, less-than-pleasant way. The ongoing COVID-19 pandemic made itnecessary for this year´s conference not to take place as usual: colleagues could notmeet for personal exchange and there was no opportunity to get to know a foreign cityand enjoy local food. This was all the more regrettable as Vienna is one of the world´smost attractive conference venues. COVID-19 also meant that many of us were bur-dened with additional obligations. We therefore considered it appropriate to extend thedeadline for the submission of contributions. Unfortunately, this put increased timepressure on the review process. Nevertheless, we are glad that in the end the reviewswere received on time. Thefirst-time organization of the ER as a virtual conference was associated with anumber of challenges. For example, organizing the program proved to be difficultbecause it was almost impossible tofind a schedule that would accommodate the manytime zones in which the participants would be located during the conference. We wereforced to make compromises here, which led to considerable limitations for individualtime zones. We regret this very much and hope for the understanding of those con-cerned. In addition, it was not possible to foresee the impact that virtualization wouldhave on the number of submissions. We are glad that the response to the call wasconsiderable despite the crisis. A total of 143 contributions were submitted, of which28 were accepted as regular papers and 16 as short papers. The papers cover a broadspectrum of innovative topics, thus underlining the great importance and attractivenessof research on conceptual modeling.We hope that the papers willfind your interest and wish you an inspiring read.Finally, we would like to thank the authors, whose contributions made the conferencepossible, the many reviewers for their outstanding commitment in preparing more than400 expert opinions, and last but not least the senior editors, without whose support wewould not have been able to cope with the evaluation of the expert opinions.

Keywords:

Keywords:
Astract: Although Model-Based Software Engineering (MBE) is a widely accepted Software Engineering (SE) discipline, no agreed-upon core set of concepts and practices (i.e., a Body of Knowledge) has been defined for it yet. With the goals of characterizing the contents of the MBE discipline, promoting a global consistent view of it, clarifying its scope with regard to other SE disciplines, and defining a foundation for the development of educational curricula on MBE, this paper proposes the contents for a Body of Knowledge for MBE. We also describe the methodology that we have used to come up with the proposed list of contents, as well as the results of a survey study that we conducted to sound out the opinion of the community on the importance of the proposed topics and their level of coverage in the existing SE curricula.

Keywords: classification modelling; request for tender; RFT
Astract: In the tender process, the customer publishes a request for tender (RFT) document containing a large list of contractly binding requirements. Suppliers need to process all of them and come up with solutions for each requirement. This thesis is written in cooperation with an industry partner on the supplier side. Since not a single person can answer all requirements, these are further assigned to responsible experts. This split is performed based on roles within the project, such as project management or technical experts for some of the companys products. Within this thesis, such a role is abstractly called subsystem. This assignment is done manually by a single person, making this task tedious and time-consuming. To support the partner, a machine learning approach is developed to automatically assign requirements to subsystems. In a literature review, suitable machine learning methods are identified, which are then compared in a benchmark to find the best configuration for each of four selected subsystems. These configurations are then checked upon generalization by evaluating them on five additional subsystems. The reasons for false classification are then identified in an interview with the person, who is currently in charge with the assignment.