In this work, we develop a system that analyzes unstructured financial news using text classification in order to forecast stock price trends. We review similar systems to build on successful ideas and combine them with novel approaches. We discuss the different types of news that are potentially relevant to the stock prices and choose news sources for the system accordingly. To eliminate irrelevant news, we present suitable filtering approaches such as the implementation of a rule-based thesaurus. We develop an automatic labeling approach and compare it to a manual labeling approach. We evaluate the influence of different automatic labeling approaches on the prediction performance. In a data training phase, we introduce a set of features novel with respect to the price forecasting task. We compare different text mining techniques such as the feature vector dimensionality reduction and different classifiers. Finally, we investigate the influence of trading costs on potential profits and run a market simulation that is able to support or reject the practical profitability of the system.