Detecting tweet sentiment and sarcasm on online learning during covid-19

Abstract

COVID-19 pandemic has created a lot of challenges for student learning and educa tion across the globe. As a result of the global increase of the state of COVID-19, numerous educational institutions in the whole world were closed in 2020 and moved to online or remote learning, which had a variety of effects on student learning. As a result teachers and students spent more time online than ever before, with both groups studying, learning, and getting acquainted themselves with information, as sets, tools, and structures in order to adapt to online or virtual learning. On the basis of COVID-19, studying different opinions about online learning as the Big Data mining and analysis of tweets from the people of various countries around the world provides the opportunity to identify, quantify, and investigate the needs, chal lenges, and interests related to online learning in various countries around the world. Analyzing the sentiment of people they want to express through their tweets gives us a clear view of their opinion about online learning. Moreover, a huge number of tweets are sarcastic and it will not be possible to crack the sentiment of a maxi mum number of people without identifying the sarcastic tweets. Different types of methods were used for these analyses. Twitter is the most popular and used social media platform around the globe for many years. So, tweet data in the form of search interests related to online learning was mined for the creation of this dataset using Rapid Miner and Twitter API. As the dataset is created based on only the tweets during the covid19 the data is much less to get a more perfect result.We have analyzed the sentiment using the stemmed feature and applied a few models among which we get the best result from the logistic regression model which is 70.63% and for the sarcasm detection, we used 3 features and overall get the best accuracy 76.19% from tf-idf, 76.94% from the stemmed feature. The more information the datasets can have the more identical the changes will be. So, work on the datasets should also be continued.