An approach for recommending relevant articles in news portal based on Doc2Vec

News portals are among the most popular websites, and their main goal is to bring the latest news to their readers. Also, it is important to provide relevant content to various types of readers. In this article, we propose an approach for recommending relevant articles on the news portal based on the content of a specific article. The proposed approach is based on Doc2Vec. The main steps of the proposed approach and training of the Doc2Vec model are described. The article also deals with text similarity problems and limitations of the Czech language in the context of recommending relevant articles. For experiment verification of our approach, random articles from the selected news portal were selected. For each article, our approach recommends the most relevant similar articles. Then, the relevant and irrelevant articles were marked. And finally, the ratio of proposed relevant articles for each random article was calculated. The experimental results show the accuracy and relevancy of the proposed approach.