Fine-grained sentiment analysis on financial microblogs and news headlines

Sentiment analysis in the financial domain is quickly becoming a prominent research topic as it provides a powerful method to predict market dynamics. In this work, we leverage advances in Semantic Web area to develop a fine-grained approach to predict real-valued sentiment scores. We compare several classifiers trained on two different datasets. The first dataset consists of microblog messages focusing on stock market events, while the second one consists of financially relevant news headlines crawled from different sources on the Internet. We test our approach using several feature sets including lexical features, semantic features and a combination of lexical and semantic features. Experimental results show that the proposed approach allows achieving an accuracy level of more than 72 %.