Automatic detection of linguistic defects in business process descriptions

This research aims to demonstrate whether it is possible to improve the quality of textual descriptions of business processes by automatic detection of linguistic defects in these descriptions. Linguistic defects are word or sentence features which pose an obstacle to the intended understanding of a text.In order to demonstrate this thesis, a design-science research approach is followed. Existing natural language processing techniques are used to develop an algorithm for the detection of one particular defect type, namely "actor unclear". This type has been found both to occur relatively often and to represent most common characteristics of linguistic defects in general. After development of the algorithm as an artifact, it is tested based on an evaluation pattern which simulates costs for positive and for falsenegative hits. These modeled costs are used to compare the performance of the artifact against several baselines. The results shall prove whether or not the use of natural language processing techniques may contribute to efficient quality improvement for textual descriptions of business processes.