Aspect Extraction from Bangla Reviews using Convolutional Neural Network

The extensive customer reviews in web assist customers for purchase-decision-making as well as providers for business planning. Summarization of reviews is desirable as reading all reviews is not feasible to evaluate properly. To find aspect categories from reviews is a sub-task of summarization known as Aspect Based Sentiment Analysis (ABSA). In this paper, we present two Bangla datasets to perform ABSA task. We collected user comments on cricket game and annotated manually. The other dataset consists of consumer reviews of restaurants. A model to extract aspect category based on Convo-lutional Neural Network (CNN) is presented. The model shows the convincing performance of the proposed datasets compared to the conventional classifiers.