Enhanced automated text categorization via Aquila optimizer with deep learning for Arabic news articles

Abstract

Text Classification is the traditional Natural Language Processing (NLP) task. Text classification (also known as categorization) has become a cutting-edge research area in recent years. However, this task has received less attention in Arabic due to the need for more extensive resources for training Arabic text classifiers. In the area of text classification for Arabic news articles, deep learning (DL) methods, namely recurrent neural network (RNN) and convolutional neural network (CNN), were effectively used. This model is trained on labelled datasets around many news topics to automatically categorize articles into predetermined classes. These DL techniques can efficiently discern the subject matter by leveraging the contextual and semantic data embedded in the Arabic text, enabling accurate classification. This application of DL facilitates effective retrieval and organization of Arabic news articles, which supports tasks such as personalized content recommendations, information retrieval, and summarization. Therefore, this study presents an Enhanced Automated Text Categorization via Aquila Optimizer with Deep Learning for Arabic News Articles (TCAODL-ANA) technique. The TCAODL-ANA technique aims to detect and classify Arabic news articles into seven classes. The TCAODL-ANA technique follows pre-processing and the FastText word embedding process to accomplish this. In addition, the TCAODL-ANA technique utilizes an effective attention-based bidirectional gated recurrent unit (ABiGRU) method to identify various news articles. To enhance the detection results of the ABiGRU method, the AO model is employed for the hyperparameter selection process. A comprehensive simulation evaluation is performed to emphasize the improved performance of the TCAODL-ANA technique. The investigational validation portrayed the superior outcomes of the TCAODL-ANA technique over existing techniques.