A two level learning model for authorship authentication.

Area of Science:

• Computational linguistics
• Digital forensics
• Machine learning applications

Background:

• The rise of internet usage necessitates robust methods for forensic authorship authentication.
• Existing authorship identification techniques can be enhanced by integrating diverse feature sets.

Purpose of the Study:

• To present a novel two-level learning technique for improved forensic authorship authentication.
• To enhance the efficiency of authorship identification by incorporating linguistic, statistical, and vocabulary features.

Main Methods:

• A two-level classifier was developed for authorship authentication.
• The first level utilized vocabulary features (word frequency).
• The second level employed a learning technique incorporating lexical, statistical, and linguistic features.

Main Results:

• The proposed two-level classifier achieved high accuracy in authorship identification, ranging from 94% to 96.16%.
• New features were proposed to effectively capture author writing styles.
• The methodology, tested on Arabic, is generalizable to other languages.

Conclusions:

• The developed two-level learning technique significantly improves forensic authorship authentication accuracy.
• Integrating multiple feature types (lexical, statistical, linguistic, vocabulary) is crucial for robust authorship identification.
• The proposed method offers a generalizable approach for authorship authentication across languages.