Topic Modeling, Sentiment Analysis and Text Summarization for Analyzing News Headlines and Articles

Newspapers and News Websites have become a part and a crucial medium in society. They provide information regarding the events that are happening around and how society is getting influenced by these events. For example, a pandemic like Covid-19 has raised the importance of these mediums. They have been giving detailed news to society on a variety of topics, such as how to detect the strains of the coronavirus, reasons for lockdown along with what are the other restrictions to be followed during the pandemic. They also provided information about the government policies which were built to be taken care of in case of pandemics and so on and they kept updated with the details about the development of the vaccines. Due to this lot of information on Covid-19 is generated. Examining the different topics/themes/issues and the sentiments expressed by different countries will aid in the understanding of the covid-19. This paper discusses the various models which were built to identify the topics, sentiments, and summarization of news headlines and articles regarding Covid-19. The proposed topic model has achieved a Silhouette score of 0.6407036, 0.6645274, 0.6262914, and 0.6234863 for 4 countries like South Korea, Japan, the UK, India on the news articles dataset, and it was found that the United Kingdom was the worst-hit, and it had the largest percentage of negative sentiments. The proposed XlNet sentiment classification model obtained a validation accuracy of 93.75%.