Cybercrime and authorship detection in very short texts

Abstract

The aim of the study is to investigate cybercrime and authorship detection in very short texts via a quantitative morpho-lexical approach. Results indicate that the classification accuracy based on the proposed system (using letter pair combinations as well as distinctive lexical features) is around 76%. In conclusion, the use of the self-organizing map (SOM) led to better authorship performance for its capacity to integrate two different linguistic levels (i.e. the morphological and lexical features) of each author together, unlike other clustering systems