Multi-view Learning for Semi-supervised Sentiment Classification

Standard supervised approach to sentiment classification requires a large amount of manually labeled data which is costly and time-consuming to obtain. To tackle this problem, we propose a novel semi-supervised learning method based on multi-view learning. The main idea of our approach is generate multiple views by exploiting both feature partition and language translation strategies and then standard co-training algorithm is applied to perform multi-view learning for semi-supervised sentiment classification. Empirical study across four domains demonstrates the effectiveness of our approach.