Sentiment Analysis of Arabic Tweets Related to COVID-19 Using Deep Neural Network

Abstract

Along with the Coronavirus pandemic, several other severe crises also spiraled worldwide. Different industries are getting irreparable scathed and many organizations succumbed to this havoc. There is an inevitable need to analyze different trends going on social media platforms to alleviate the fear and misconceptions among public. The research plays out a thorough investigation on the emotional directions of the Arabic public dependent on social media using Twitter platform particular. We have extracted data from Twitter from November 2020 to January 2021.There are tweets from different cities of Arab. Natural language processing NLP and Machine learning ML capabilities are used to analyze whether an opinion's sentiment is positive, negative, or neutral. This research scrapes around Arabic tweets and then after manual annotation to classify the tweets into different sentiments like negative, positive, neutral, etc. This research use TFIDF and word embedding as a feature vector and then use Long Short-Term Memory and Naïve Bayes as classification. This work using two advanced machine learning methods, present a learned long short term memory LSTM model and a Nave Bayes model on the collected tweets. In addition, compare the performance of the Nave Bayes and LSTM models. In comparison with the Naïve Bayes the LSTM model performs better with an accuracy of 99%. The work analysis helps different Government and private organizations to understand public sentiments, their behavior towards this pandemic and then act make strategic decisions accordingly. In addition, this research focuses on data visualization by displaying a sentiment plot and a word cloud.