Distilling Business Value from COVID - 19 Public Media Dataset with Machine Learning and Natural Language Processing

While manual analysis of news coverage is difficult and time consuming, methods in natural language processing can be used to uncover otherwise hidden semantics. This work analyses more than 370,000 news articles to explore connections and trends in business decisions and their financial impact during the COVID-19 pandemic. Topic modelling, sentiment analysis and named entity recognition methods are used to identify connections between the articles and the financial performance of selected companies or industries. This report sets out the results of the individual natural language processing methods and the resulting analysis with financial data. Interesting contrasting topics in the media can be filtered out that are associated with the companies with the highest or lowest positive sentiment. This information could be useful to companies to gain an understanding of topics that are currently treated favourably or unfavourably by the media and hence assist with communication strategies and competitive intelligence.