Pattern based information retrieval approach to discover extremist information on the internet

Abstract

This paper is devoted to the research and development of machine learning methods aimed at discovering potentially dangerous extremist information in social networks using pattern based approach. In this approach, a text document containing extremist information is used for automatic extracting keywords to query a social networks search engine, and then found messages are filtered according to the topic based measure of relevance with the pattern. N-gram based algorithms are proposed for constructing hidden topics and keywords that allow applying the proposed approach in the case of multilingual and illiterate texts. The performance of the proposed methods is experimentally studied on benchmark Ansar1 dataset.