Towards Automated Fiqh School Authorship Attribution

ورقة مؤتمر

2022

19th International Conference on Computational Linguistics and Intelligent Text Processing

Abstract. The word Fiqh (Islamic jurisprudence) refers to the body of Islamic law (Shari’ah). A large volume of Fiqh literature has been generated over the past thirteen hundred years, some of which texts have unknown authors. The importance of identifying the Fiqh School emanates from its importance in offering an authenticated interpretation of fundamental sources.
The traditional method for identifying the Fiqh School for a certain text is either by knowledge of the school affiliation the author or by close reading of the text by Fiqh scholars. This method is costly in terms of the time and human effort involved. An alternative to this manual approach is automated identification of Fiqh school texts using stylometric analysis. In this study we investigate the ex- tent to which stylometric features can be used as predictors for Fiqh school authorship of a given text. We explore a corpus of Arabic Fiqh texts using unsupervised cluster analysis and supervised machine learning.
The results of our study show that the Fiqh schools have distinctive text style features that can be used to indicate authorship. The observations from the cluster analysis experiment using a number of different distance measures are visualized using network graphs. The best clustering in terms of Fiqh school division was achieved by the Classic Delta distance measure and Eder’s Delta distance measure. The results from the supervised experiment comparing the four classification algorithms: Support Vector Machines (SVM), Naïve Bayes (NB), K-Nearest Neighbor (KNN), and Delta show that supervised classification using SVM produces the highest average accuracy at 97.5% for the task of Fiqh school prediction.

Keywords: Fiqh school attribution, Stylometric Analysis, Arabic language, Cluster analysis, Supervised Classification

• نوع عمل المنشور: Original Article (EDP 2022)