Based on a collaborative data science environment, and a large document corpus (> 130.000 documents from German tax law) we demonstrate the extraction of semantic information. This paper shows the potential of rule-based text analysis to automatically extract semantic information, such as the year of dispute in cases. Additionally, it demonstrates the extraction of legal definitions in laws and the usage of terms in a defining context. Based on an iterative and interdisciplinary process, involving legal experts, software engineers, and data scientists, to evaluate and continuously refine the model used for the computer-supported extraction.