Density based active self-training for cross-lingual sentiment classification

Cross-lingual sentiment classification aims to utilize annotated sentiment resources in one language (typically English) for sentiment classification in another language. Most existing research works rely on automatic machine translation services to directly project information from one language to another. However, since machine translation quality is still far from satisfactory and also term distribution across languages may be dissimilar, these techniques cannot reach the performance of monolingual approaches. To overcome these limitations, we propose a novel learning model based on active learning and self-training to incorporate unlabeled data from the target language into the learning process. Further, in this model, we consider the density of unlabeled data to avoid outlier selection in active learning. The proposed model was applied to book review datasets in two different languages. Experiments showed that the proposed model could effectively reduce labeling efforts in comparison with some baseline methods.