Classification of Persian Product Reviews Using Neural Networks

With the rising influence of reviews on online retail shopping, automated opinion mining of consumer reviews is becoming increasingly important. Opinion mining or the classification of reviews is done using machine learning algorithms or neural networks, yet works in this area for the Persian language are limited. This paper tries to implement and demonstrate the performance of three neural networks by training them on the product reviews dataset from DigiKala, an Iranian e-commerce company. An LSTM and a BiLSTM model are used, and an RNN is used as the baseline to show how effective the aforementioned models are at classifying the samples, which the BiLSTM model shows slightly better results than the other two