LSTM Algorithm with FastText Word Embedding Model for Evaluating Arabic Sentiment Analysis

With the rise of social networks, online users increasingly express their sentiments on various topics. Sentiment Analysis (SA), an essential area in Natural Language Processing (NLP), aims to identify the polarity of senti-ment and derive insights from public opinions. The Arabic language poses significant challenges for SA because of its diverse dialects, complex morphology, and syntax. Neural Networks (NN) models in Deep Learning (DL) are highly effective for sentiment classification in many tasks, especially in the education sector thanks to their advanced abilities to analyze textual data. Recurrent Neural Networks (RNNs), particularly Long Short-Term Memory (LSTM), networks, have demonstrated remarkable proficiency in understanding sequential data for SA tasks. Besides they are skilled at capturing context and trends in longer texts. These capabilities enhance sentiment analysis accuracy, providing educators with valuable insights into student satisfaction and course effectiveness. This paper offers an evaluation study of Sentiment Analysis for the Arabic language (ASA) employing an LSTM approach alongside the FastText word embedding model. Our experimental results confirm that the LSTM model with FastText enhances text classification accuracy in both datasets, with the first dataset achieving higher accuracy than the second.