This paper investigates how stock-specific and market-wide news sentiments, obtained from Thomson Reuters News Analytics, affect abnormal returns of S&P 500 stocks. It is well-known that the relationships between the stock-specific news sentiment and raw stock returns are rather weak. This can be explained by the fact that stock returns are driven predominantly by the market factor as well as some other well-known fundamental factors, and in a lesser extent by the idiosyncratic stock-specific information. Using factor models of Fama and French and of Carhart, we remove the influence of the fundamental factors from S&P500 stock returns. This allows us to investigate the relationships between the news sentiment and abnormal, i.e., idiosyncratic component of returns. We use both stock-specific and constructed market-wide sentiments for this purpose.We find that abnormal returns show a strong relationship with the news sentiment, which is consistent across sectors. Moreover, we find that the market-wide news sentiment significantly amplifies the effect of stock-specific news. Furthermore, we investigate separately the effect of news on various sectors, on small, medium and large stocks (in terms of size and book-to-market) and on less and more volatile stocks, and find that there are significant deviations on how abnormal returns react to positive and negative sentiment in news. Since the factor models are fundamentally tradable, our findings can be used to create profitable trading strategies.