Text reuse detection by keyword extraction for telegram channels

Text reuse detection is the task of finding similar texts, which has many applications e.g. in plagiarism detection and analysis of information diffusion. The general approach to this problem is by detection of texts similarities in combination with other features such as time-stamp which can be used to specify the precedence of publishers e.g. to find the first publisher. In this article a method for finding similar texts has been proposed using keyword extraction which operates in linear time like LSH1 method. In addition, it supports dynamic inputs and does not depend on text vector dimensions. Our evaluations show, it has better performance in clustering quality measures and also run time.