The purpose of this research is to contribute to the academic field by demonstrating the relationship between stock related Twitter messages, their frequencies, sentiment analysis; stock return, volume, and volatility of Dow Jones Index and BIST30 & BIST100 Index. In this study, The Multinomial Naive Bayes Text Classifier is used as methodology since it is the most conventional method for text classification based on previous research. Using computational linguistics methods, 138.070 English and 34.632 Turkish tweets have been analyzed on a daily basis for a period of 8 months. The results demonstrated a strong relationship between tweets and their impact on the market. Moreover, according to results, there is a positive correlation between the number of retweets and BIST Volume lag-1 and lag+1. In addition, this article confirms that stock microblogs contain valuable information for investors and it can be an assistance in predicting the future market index.