Hybrid CNNs-LSTM Deep Analyzer for Arabic Opinion Mining

Deep learning models have showed great capabilities in data modelling on natural language processing various applications, including sentiment analysis, part-of-speech tagging, machine translation, and many others. In particular, convolutional neural network (CNNs) and long-term short memory (LSTM) have proved to be effective in capturing long-term dependencies in sequential data that result in state-of-the-art performance in comparison to traditional machine learning algorithms. This research paper, therefore, structures an enhanced model of both CNNs and LSTM for the feature resourcefulness of Arabic text data on freely available benchmark datasets, with word2vec representation model for each corpus. The model is projected for Arabic sentiment analysis (ASA) in highlight. The proposed architecture has achieved better performance on three datasets out of five in comparison to previous studies. In research conduct, the model achieved a total accuracy of 0.881 for Main-AHS, 0.968 for Sub-AHS, 0.842 for Ar-Twitter, 0.7918 for ASTD, 0.903 for OCLAR.