Deep-Learning Based Aspect Sentiment Classification for Low-Resource Language

Task of sentiment analysis is being increasingly recognized as an important one, not just from an academic but also from a business perspective. The bulk of the approaches that are now in use, make an effort to identify the overall polarity of a sentence, paragraph or text span, regardless of the entities that are addressed in the review. Aspect-based sentiment analysis (ABSA) deals with sentiment at a fine level. In this paper, a deep learning based model of aspect sentiment classification (ASC) for low-resource language data is proposed. It is based on multiclass classification paradigm. The presented model employs a BiLSTM unit followed by a dense layer. Experimental results have yielded accuracy 72.03% for well-accepted IIT Patna Hindi dataset. This is higher than that of previously reported works for Hindi.