TU München - Fakultät für
Es sprechen Studenten über ihre abgeschlossenen Diplomarbeiten und Systementwicklungsprojekte.
Am Mittwoch, 04.03.20, ab 10:15 Uhr, im Raum Alonzo Church (01.09.014):
Model Generation from Text for Practical Requirements
In the final test phase of a software system the developed system has to be tested according to acceptance tests. For this purpose the specification created at the beginning is used to test the system for validity. The procedure for this is to manually check the specification and create test cases for it. Since this procedure has to be conducted by humans it is time consuming and therefore cost intensive as well as error-prone. The potential which is already provided by natural language processing tools is still widely unused. Therefore, the already existing approach to extract models from user stories should be evolved further. For this reason first a case study is conducted in this thesis for a detailed analysis of the content structure. With this result of the case study the aforementioned approach is further developed in view of coreferences. Then an analysis for the natural language processing tool CoreNLP with coreferences is made and the model extraction approach is refined. A model-to-model transformation is implemented, which is based on combining already extracted models with respect to coreferences. Conclusively, a superior and more abstract entity model is defined to integrate these models in a suitable structure for a better classification.