Social media data is unstructured data where these big data are exponentially increasing day to day in many different disciplines. Analysis and understanding the semantics of these data are a big challenge due to its variety and huge volume. To address this gap, unstructured Arabic texts have been studied in this work owing to their abundant appearance in social media Web sites. This work addresses the difficulty of handling unstructured social media texts, particularly when the data at hand is very limited. This intelligent data augmentation technique that handles the problem of less availability of data are used. This article has proposed a novel architecture for hand Arabic words classification and understands based on convolutional neural networks (CNNs) and recurrent neural networks. Moreover, the CNN technique is the most powerful for the analysis of Arabic tweets and social network analysis. The main technique used in this work is character-level CNN and a recurrent neural network stacked on top of one another as the classification architecture. These two techniques give 95% accuracy in the Arabic texts dataset.

News published on social networks has a notable impact on changing people's perceptions on various topics. However, all news available on social media may not be genuine and might come from unverified sources. The prevalence of fake news is an inevitable concern that needs to be addressed effectively. This study presents an ensemble algorithm to improve fake news detection tools. Long-Short-Term-Memory (LSTM) and an ensemble of LSTM and Convolutional Neural Networks (CNN) were used. The proposed model used bidirectional LSTM layers and CNN convolutional 2D layers with kernel sizes of 2, 3, and 4 for 2-gram, 3-gram, and 4-gram tokens. The results obtained show an accuracy of 96.7% and 97.3% on a fake news dataset using the LSTM model and CNN-LSTM model, respectively, significantly improved from the maximum accuracy of 94.88% reported in a previous study. Embedding layers yielded significant improvements when paired with extended word sequences and pre-trained embedding vectors. Diverse tokenization methods with and without pre-trained embedding layers were also considered. The ensemble model achieved a 10.03% improvement in predictive accuracy on the Liar dataset, compared to the 6.08% improvement reported in a previous study using the same dataset.