Michael Arens, Lea Djomo, Franck Djeutang, Oliver Schmidtke, Adrian Graap, Dmitrii Romanov - Data Science (FB5) Semesterarbeit


As part of the AdvoInformer research project, a dialog-based application (a chatbot with question-answering functionalities) is to be developed that can answer questions about occupational health and safety. As part of the corresponding (larger) research project AdvoInformer, software for automated risk assessment according to the German Occupational Health and Safety Act (German law: §§ 5,6 ArbSchG) is to be created. In concrete terms, this means for this student project that a procedure for text analysis is to be developed, which analyzes corresponding texts (legal text) and whose contents are subsequently to be retrievable via Question Answering in a dialog-based manner. This system is to be developed here exemplarily for a limited set of specific legal texts from the area of employment protection provisions (German: Arbeitsschutz). The goal is to obtain answers to questions formulated in natural language. In order to obtain corresponding results, it is necessary to build up an appropriate data structure. For the implementation of the project, a set of relevant German and English documents will be made available.


The team of students successfully implemented a chatbot that is able to retrieve a corresponding paragraph from legal unstructured documents in order to answer a question of a user.