| Abstrakt: |
Companies with diverse product offerings rely on customer reviews to gauge product reception. Following a purchase, customers often share their opinions on the website. Prospective buyers, prior to deciding, typically peruse these reviews to inform their choices. Analysing such feedback, whether positive or negative, holds paramount importance for companies seeking to improve product quality. Researchers are actively exploring methods to categorize comments based on sentiment scores. Notably, customers may express their reviews in Arabic text. Despite challenges such as the structure and morphology of Arabic text, a scarcity of machine-readable Arabic dictionaries, and limited tools for handling Arabic text, minimal progress has been made in the analysis of Arabic reviews. While some attempts have been undertaken, they have achieved suboptimal accuracy. In response, the authors propose a hybrid deep learning model comprising Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) with GlobalMaxPooling. Through multiple iterations, the authors fine-tuned the proposed model and applied it to publicly available Arabic Reviews dataset, achieving a notable 95% accuracy, precision, recall, and F1 score. The results indicate that, when compared to alternative models, the proposed model exhibits superior accuracy. [ABSTRACT FROM AUTHOR] |
