Fakultät für Informatik

TU München - Fakultät für Informatik
Software- and Systems Engineering Research Group



Es sprechen Studenten über ihre abgeschlossenen Diplomarbeiten und Systementwicklungsprojekte.

Am Montag, 29.10.18, ab 10:00 Uhr, im Raum Neumann (00.11.038):

10:00 - 10:25:Arjen SpaansBA (Henning Femmer, Sebastian Eder)Automatic Detection of Requirements Smells in User Stories
10:25 - 10:50:Marcel BrucknerBA (Elmar Jürgens)Recognition of Generated Code in Open Source Software

Automatic Detection of Requirements Smells in User Stories

The requirements within agile software development methodologies are often documented in the user story format. The quality of these user stories is regarded as crucial for the success of agile software development projects. So far, we have seen limited research regarding user story quality. Also, the limited research is reflected by the small number of automated quality detection tools specially designed for user stories. We aim to address these problems in this thesis by defining and developing a new quality model and automated detection tool regarding user story quality. We address these problems in three steps. Step one starts with consolidating existing user story quality factors found in literature, such as the INVEST quality model. We consolidate those existing quality factors in an Activity-Based User Story Quality Model (ABUS-QM). Based on the ABUS-QM, we develop an automated quality detection prototype. The prototype is a software solution that uses common NLP-Technologies for the quality factor detection heuristics. The second step focuses on gaining more insight into the accuracy of quality factor detection heuristics. We classify the quality factors by their automated detection accuracy. This classification establishes well-founded expectations of our prototype and future automated user story quality detection solutions. The evaluation of our prototype against the accuracy classification is the last step in our approach. We let user story experts evaluate the results of several product backlog reviews done by our prototype. Also, we compare the results of automated and manual reviews. The results of this evaluation give us insight into the recall and precision characteristics of our prototype. We demonstrate that our automated quality detection prototype is capable of detecting four quality factors. The results of the prototype evaluations show medium recall and precision characteristics. Our ABUS-QM contributes to the current understanding of user story quality. With the ABUS-QM we provide a more precise quality specification in the context of an activity. The detection accuracy classification of the ABUS-QMs quality factors establishes goals for future automated detection solutions. Also, our prototype shows the potential of common NLP-technologies in the context of analyzing user story quality.

Recognition of Generated Code in Open Source Software

A suffix-tree clone-detection approach to find clones among the comments in source code files is presented. Using this approach we found a multitude of generator-patterns that are inserted by their respective code generator and identify the source code file as generated. We extracted the patterns to build up a generator-pattern repository that can be used to automatically classify code into the categories generated and manually maintained. This repository is used on a reference data set and a huge, randomly composed collection of open source projects to test its capabilities and to calculate different proportions of generated code.

© Software & Systems Engineering Research Group
Sitemap |  Kontakt/Impressum
Letzte Änderung: 2018-10-22 14:13:07