In most cases, voters only have a vague and subjective perception of a newspaper’s proximity to political parties. If voters are uninformed about the political attitude of media reporting, they may be manipulated in their democratic opinion-forming. One answer to biased news and false information is transparency and quantifiability. For this reason, we introduce the Sentiment Political Compass, a data-driven framework to analyze newspapers with respect to their political conviction. Newspapers are embedded in a two-dimensional space (left vs. right, libertarian vs. autocratic) resembling a compass, which serves as a tool for analysing relative political proximity. We provide technical details of the system, including the framework to crawl newspaper articles, locate and extract entities and perform entity sentiment analysis. We demonstrate the analytical power and informative value of our approach by analyzing over 180,000 newspaper articles with over 740,000 sentiments surrounding the federal elections 2017 in Germany. Since our model can be reproduced entirely open-source, it may be applied to classify the political landscape in any country in the world.