Data Analysis of the Web News Headlines based on Natural Language Processing

Abstract

— This paper explores the problem of media content data analysis with the focus on the phenomenon of vaccination, closely related to the COVID-19 pandemic. The presented research is an extension of the previous work, but it differs in two main areas. Firstly, the text corpus submitted to the analysis has been considerably increased. Secondly, the previous data analysis was performed on the body part of the posts, while now it is focused on the most prominent part of the news posts, their headlines. This change from body to headline analysis was provoked by significant differences in their characteristics and the fact that most people read only headlines. Described data acquisition uses an advanced content collection approach followed by the modeling process, during which a set of natural language processing algorithms were applied. To enable the comparison, the model uses the same set of algorithms in the modeling phase like in previous work. The main contributions of the work are manifested in: i) approaching the problem from a new perspective, ii) applying more efficient method of data collection, and crucially iii) enabling the comparison of analysis results for individual parts of the content, which ensured a comprehensive insight into the characteristics of news posts.