Sentimental Analysis of Arabic Tweets related to COVID-19 using AraBERT model

Coronavirus has affected the lives of almost everyonebelonging from every aspect of life. COVID-19 has caused depressions, anxiety, stress, fear among the people. Every countryaround the world is facing various challenges amid COVID-19 and trying to implement some measure in order to control the spread of COVID-19. Now-a-days, social media is a powerful tool and people use it for sharing their feelings and thoughts. Like people of all other counties and cultures, Arab people also sharedtheir thoughts on social media during COVID-19. However, the Arabic language is rich in morphology and has a large number of dialects. Arabic Sentimental Analysis has attracted many researchers but still the research work on Arabic Language is limited. In this research, we used AraCOV-19 data, an Arabic language dataset related to COVID, and assigned the sentiments labels to each tweets using Textblob-ar, after doing necessary pre-processing. We classify the negative and positive tweets using AraBERT model and compared the results with state-of-the- art models. The results showed that our model outperformedall other state-of-the-art models by achieving 93% and 92% accuracy for positive and negative classification respectively. Hence, such methods of sentimental analysis are very helpful in devising new policies for the control and prevention of epidemics like COVID-19.