Extracting character networks is an initial step towards the difficult task of automatically analyzing novels. This extraction process consists of different steps like recognition of entities (entity recognition), refrences and mentions of the same character (coreference resolution), attributes of characters and relationships between characters as well as their nature (family, love, social and other). Because every step of this process is difficult, the error propagation of a pipeline handling this task is very high. The goal of this project is improving all steps of this process in order to extract expressive character networks for different genres (e.g. fairy tales) of novels.