Arabic Sentiment Analysis based on Deep Reinforcement Learning
Mohamed Zouidine, Mohammed Khalil · 2022
In this work, we handle the problem of Arabic sentiment analysis by combining the Arabic language understanding transformer-based model AraBERT and an LSTM-CNN deep learning model. We propose a new training objective function based on deep reinforcement learning that combines cross-entropy loss from maximum likelihood estimation and rewards from policy gradient algorithm. We evaluate our proposed system on the LABR book reviews dataset. Experimental results show that the proposed model outperforms the state-of-the-art models and provides an accuracy of 87.58%.