Searching for semantical connections between nouns and adjectives within sentences proves to be useful in investigating the potential impacts of publicly available news on daily returns. We use extraction of composite expressions of adjective-noun pairs from economic and financial news based on a sentence based text processor. Analysis of 10 year period using seven news journals and daily closing stock prices results in approximately 70% precision and 50% recall indicators on S&P500 stocks related to trading days when extreme positive or negative returns were realized by the majority of stocks. Indicators of general economic news are better then firm specific headline news. News on days T+1 corresponds with returns better than news on day T or T-1. Our method outperforms the analyzed bag of words alternatives.