Thematic Clustering Methods Applied to News Texts Analysis

This paper is devoted to a problem of partition documents from the news flow into groups, where each group contains documents that are similar to each other. We use thematic clustering to solve this problem. The existing clustering algorithms such as k-means, minimum spanning tree and etc. are considered and analyzed. It is shown which of these algorithms give the best results working with news texts. Clustering is a powerful tool for text processing, but it can't give a complete picture of news article semantics. This paper also presents a methodic of comprehensive news texts analysis based on a combination of statistical algorithms for keywords extracting and algorithms forming the semantic coherence of text blocks. Particular attention is paid to the structural features of the news texts.