HAT - A new corpus for experimental stylometric evaluation in Arabic
DOI:
https://doi.org/10.36505/ExLing-2021/12/0052/000525Keywords:
natural language processing, computational linguistics, author style analysis, stylometry, digital librariesAbstract
This paper presents a new Arabic corpus designed for authorship attribution research. The corpus, named HAT (Hundred of Arabic Travelers), contains 300 texts extracted from 100 Arabic travel books written by 100 different authors. Each author is represented by three separate texts. The corpus was developed to address the shortage of suitable Arabic datasets for stylometric evaluation. Baseline experiments using an SVM classifier and Leave-One-Out cross-validation showed promising results, achieving an authorship attribution accuracy of up to 95% using character tetra-grams. The HAT corpus is proposed as a freely available reference dataset for Arabic author style analysis and experimental evaluation.
References
Eder, M. (2010). Does size matter? Authorship attribution, short samples, big problem. In Digital Humanities 2010 Conference (pp. 132-135). London.
Forest, D. (2006). Application de Techniques de Forage de Textes de Nature Prédictive et Exploratoire à des Fins de Gestion et d’Analyse Thématique de Documents Textuelles Non Structurés (PhD thesis). Université du Québec à Montréal.
Juola, P. (2006). Authorship attribution. Foundations and Trends in Information Retrieval, 233-334.
Küppers, R., & Conrad, S. (2012). A set-based approach to plagiarism detection. In PAN 2012 Lab: Uncovering Plagiarism, Authorship, and Social Software Misuse, held in conjunction with CLEF 2012, Rome, Italy, 17-20 September 2012.
Sayoud, H. (2012). Author discrimination between the Holy Quran and Prophet’s statements. Literary and Linguistic Computing, 27(4), 427-444.
Tambouratzis, G., Markantonatou, S., Hairetakis, N., Vassiliou, M., Carayannis, G., & Tambouratzis, D. (2004). Discriminating the registers and styles in the Modern Greek language – Part 1. Literary and Linguistic Computing, 19(2), 197-220.
Downloads
Published
Issue
Section
License
Articles are published under the Creative Commons Attribution 4.0 International License (CC BY 4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.
