Using Text Mining Techniques to Understand the Economic Effects of COVID-19 Pandemic

Abstract

Purpose: This study aims to analyze the headlines of news in Turkish using text mining to determine the effects of COVID-19 on the economic situation. Approach/Methodology/Design: The headlines of the economic news in Turkish, were obtained from online sources. Tokenization was applied to convert headlines to words. The words accepted as stop words such as 'and', 'the', 'most', 'happened' were removed from the dataset. Term-document matrix was created. The most frequently used three words or phrases were determined as 'increased', 'decreased', and 'in the USA', according to the term-document matrix. Separate datasets were formed for each word or phrase. The news in each dataset was labelled manually by examining and interpreting in what sense they were positive, negative, or neutral. Finally, labelled news analyzed to determine sentiment of news per each month. Findings: When the findings are evaluated altogether with the economic indicators in the relevant period, they correspond to the economic indicators. Moreover, this demonstrates that an effective conclusion regarding economic developments can be made by applying text mining techniques on headlines of the economic news. Practical Implications: The authors have indicated and discussed, importance of using text mining techniques to understand economic situation. Originality/Value: The study presented that text mining techniques can be applied on economy news to understand direction, status of global and local economies and effects of COVID-10 on the economy.