Full edition for scientific use. This dataset contains a German-language sentiment dictionary of 5001 negative words and their associated sentiment strength on a five-point-scale from 0 (not negative) to 4 (very strongly negative). The procedure for building the dictionary contains the following steps: (1) Sampling 12713 sentences from Austrian parliamentary debates (1995-2013), party press releases (2006-2013), and political news reports (2013), (2) Crowdcoding the sentiment strength of sentences, (3) Estimating a sentence tonality score (4), Estimating a word tonality score, (5) Discriminating between important and unimportant words.