Abstract:

Computer science can be described as the discipline of modelling par excellence, because every form of processing knowledge and information takes place via models. The design of software is usually based on requirements. "Real-world" requirements are transferred into "software-technically realizable" concepts by means of modelling. One guiding technique therefore is a noun-verb analysis. Thus, one characteristic of a model is the reduction feature [1] which means that a model does not have all the features of the original, but only a few - and possibly also those in a modified, "similar" form. The subjective assessment of a software designer therefore plays a role in the creation of a model. We are interested in determining what the role of subjective assessment is in relation to or in conjunction with the application of noun verb analysis. Therefore, we designed an eye tracking study that investigates how class diagrams can be generated, based on a requirements specification.

The eye tracking study surveys how expert programmers in comparison to novices generate class diagrams [2]. The study has ten tasks. Each task consists of one requirement and three suitable class diagrams. The participants have to choose the most suitable diagram and give a reason for their choice. Up to now, a sample of 30 participants is recruited for the purposes of this study. This sample consists of students with limited programming skills and experts who are working for small and medium-sized enterprises and have to generate class diagrams on a daily basis.

Currently, the data is processed and analyzed and data analysis is still in progress and will be finished for the final papers. A possible outcome the study may come up with is that the noun/verb analysis on requirements written using the template of Chris Rupp [3] can help to find possible candidates for class names, methods or attributes in a class diagram. The expectation is that when analyzing a requirements specification, novices will read through the whole sentence more than once and will list all nouns as classes whereas experts will concentrate on the nouns and verbs in the requirements specification that are realizable for a software system. The first basic analyses seem to proof the idea that experts perform in a more efficient way. They can solve the tasks with lesser and shorter fixations and make better choices while they are modeling their class diagrams.

References:
[1] Hesse, W., & Mayr, H. C. (2008). Modellierung in der Softwaretechnik: eine Bestandsaufnahme. Informatik-Spektrum, 31(5), 377–393. https://doi.org/10.1007/s00287-008-0276-7
[2] K.A. Ericsson, R.T. Krampe, and C. Tesch-Römer. „The role of deliberate practice in the acquisisiton of expert performance”. Pychological Review, vol. 10, 3, pp. 363-406, 1993.
[3] C. Rupp. Requirements-Engineering und -Management: As der Praxis von klassisch bis agil. Munich, Germany: Hanser, 2014.

Keywords:

Software Engineering, Eye Tracking, Expertise, Requirement Engineering, UML.