Fine-grained Sentiment Analysis of Foreign Exchange News

Abstract

Foreign exchange (Forex) news plays a significant role in asset pricing, risk assessment and exchange rate forecasting in Forex markets. In this work, we leverage machine learning algorithms to probe the sentiment orientation and sentiment intensity of Forex news. In the aspect of sentiment orientation, two cases (text embedding to the vector and fusion of sentiment words' weights) are evaluated to investigate the sentiment orientation. Moreover, confusion matrix is adopted to further analyze the classification results. In terms of sentiment intensity, three categories of words are considered and grid search is applied to seek the weights of words. Experiments indicate that this paper has achieved relatively good results in sentiment analysis.