Multilingual Subjectivity Detection Using Deep Multiple Kernel Learning

Subjectivity detection can prevent a sentiment classifier from considering irrelevant or potentially misleading text. Since, different attributes may correspond to different opinions in the lexicon of different languages, we resort to multiple kernel learning (MKL) to simultaneously optimize the different modalities. Previous approaches to MKL for sentence classifiers are computationally slow and lack any hierarchy when grouping features into different kernels. In this paper, we consider deep recurrent convolution neural networks to reduce the dimensionality of the problem. Further, the lower layers in a deep model are abstract and the higher layers become more detailed connecting attributes to opinions. Hence, the features learned automatically in the multiple intermediate layers can be used to train MKL classifiers depending on the application. The proposed deep recurrent MKL outperforms the accuracy of baselines by over 5-30% and is several times faster on two benchmark datasets for subjectivity detection. It can also be used to develop subjectivity lexicons in other languages using English.