One can experience strong polarization in the judgement of the current economic and political events. Social media and online news portals serves as the main platform where people discuss these events. Over the last few years strong polarization has emerged between the Hungarian news portals as they more and more emphasize their beliefs in their articles. In my research I study whether the application of natural language processing techniques is able to verify this polarization. As a first step I build a corpus of Hungarian news articles by implementing Web Crawlers for archiving the articles and I apply additional processing steps to structure the corpus in a uniform manner. Afterwards I apply keyword extraction and topic modelling on the corpus in different time frames so I analyze how the keywords and abstract topics can describe that periods. Using the presence of keywords and abstract topics by portals I evaluate how well they characterize the portals and the similarities between them.