Abstract
Nowadays Twitter and Facebook are the most popular social media platforms to express opinions about various topics. Sentiment analysis (SA) aims to extract the opinions from tweets, reviews and comments. Almost all prior research has used machine learning algorithms to classify the Arabic text into positive, negative or neutral. Recently, the deep learning techniques have shown promising accuracy for SA on English, Thai, Persian, and Tamil. Motivated by the results, we propose the first study applying deep learning to classify sentiments of Saudi dialect texts. We have collected a dataset of 32063 tweets and then applied two deep learning techniques to perform SA: Long short-term memory (LSTM) and bidirectional long short-term memory (Bi-LSTM). For the sake of comparison, we have also applied the well-known support vector machine (SVM) algorithm to classify the sentiments of the collected data. The results show that the deep learning techniques outperform the SVM algorithm. The experimental result of Bi-LSTM is 94% outperforming that of the LSTM (92%), while the SVM has the lowest performance of 86.4%.