Financial Aspect and Sentiment Predictions with Deep Neural Networks: An Ensemble Approach

In this paper, we describe our ensemble approach for sentiment and aspect predictions in the financial domain for a given text. This ensemble approach uses Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs) with a ridge regression and a voting strategy for sentiment and aspect predictions, and therefore, does not rely on any handcrafted feature. Based on 5-cross validation on the released training set, the results show that CNNs overall perform better than RNNs on both tasks, and the ensemble approach can boost the performance further by leveraging different types of deep learning approaches.