AUTHORSHIP ANALYSIS IN ELECTRONIC TEXTS USING SIMILARITY COMPARISON METHOD

Abstract

The most recent changes to the criteria in legal process for scientific evidence have emphasized scientific methods of authorship analysis. This study examined the authorship of electronic texts using a quantitative method based on forensic stylistics and computer technologies. This study uses 300 digital texts produced by 100 authors, including 100 questioned texts (Q-text) and 200 known texts (K-text). Personal texts of WhatsApp messages are used in this study as electronic texts. Authorship analysis was conducted by tracing the n-gram and testing all the text sets using the Similarity Comparison Method (SCM). Based on the results of the word 1-gram test, the SCM accuracy was found to be quite high, ranging from 85% to 96%. The findings of employing the tiny set are promising, with the various stylistic traits offering dependable accuracy ranging from 92% to 98.5%. The character-level n-gram tracing indicates a key feature of authorship attribution.