Both traditional finance and behavioral finance theory have reached a consensus that the news media de facto influence stock prices to some extent. There is also evidence that investors are not only subject to the sentiment of related news articles but also the public opinions. The challenge lies on how to quantify such sentimental information to predict the movement of stock market. To measure the sentiments of articles and capture the public mood from postings, we construct and maintain a sentiment dictionary. We utilize both the official information from news articles and user postings in discussion boards to predict firm-specific stock price, and differentiate various types of news articles in the predictive model. Our experiments on CSI 100 stocks during a six week period show a predictive performance in closeness to the actual future stock price is 0.03503 in terms of mean squared error, the same direction of price movement as the future price is 67.6%. Among all seven news topic categories, restructuring news of enterprises has the best predicting performance with direction accuracy of 68.18%.