Deep Learning Based Sentiment Analysis On COVID-19 Public Reviews

Abstract

Sentiment Analysis is a classification task in order to identify public reviews about different issues like product reviews, movie reviews, restaurant reviews, political opinions, and other current issues by extracting the public reviews from Social Media, and other Micro blogging sites. As we all know Coronavirus Disease 2019 (COVID-19) is still a global issue for entire world and people are expressing their emotions, thoughts, and opinions about this issue with help of Twitter, Facebook, and other Media. In this paper we have collected public tweets from Twitter which are talked about the COVID-19 global pandemic and applied a Convolutional Neural Network with Bidirectional Long-Short Term Memory (CNN-Bi-LSTM) hybrid Deep Learning algorithm to detect the user’s outlook on this pandemic whether they have positive feelings, negative feelings, or neutral feelings. The proposed method used preprocessing techniques to clean the data and used a word embedding pre-trained model to extract word embedding for rare words in our corpus with the help of FastText and Globe pre-trained models. The CNN-Bi-LSTM hybrid model evaluated using accuracy, precision, recall, and f1 evaluation techniques. The experimental result has been shown 99.33% accuracy using CNN-Bi-LSTM with FastText pre-trained model, and 97.55% accuracy using CNN-Bi-LSTM with GloVe pre-trained model.