Topic modeling of internet news during the COVID-19 epidemic using the latent Dirichlet allocation

The aim of this thesis study is to perform the topic modeling processes and investigate the results by making Latent Dirichlet Allocation (LDA) on the internet news published during the Coronavirus Pandemic period. For these purposes, related researches were carried out on the news published in 9 different news sources and in 10 different categories on the internet, and the results were interpreted. In addition, in the thesis study; Zemberek, a Natural Language Processing method that supports the Turkish language, was used in the process of obtaining the roots of all the words in the content of each news. The emotional state of the news was revealed on a monthly basis by using VADER, which reveals the emotional state result for each category of news. The datasets used in this study were created completely new and these news from the relevant news channels were obtained by means of bots with newly developed code structures. In the study, the most common topics in each category were determined and it was tried to determine which daily topics these topics were related to periodically.