NewsCATS is an automated text categorization (ATC) prototype using a hand-made thesaurus to forecast intraday stock price trends from information contained in press releases. Due to a unique labeling approach and by carefully selecting the appropriate training data NewsCATS achieves a performance which is clearly superior to other ATC prototypes used for stock price trend forecasting. In this paper we describe the architecture, training, and testing of NewsCATS as well as the results of an extensive robustness analysis.