Hadith Grading & Machine Learning: Feasibility of Automatic Isnād-Analysis

Abstract

This study examines the feasibility of using machine-learning (ML) techniques to assist classical hadith grading by automatically analysing isnād (chains of transmission). The paper proposes a hybrid framework combining (1) robust dataset construction (graph and sequence representations of sanads), (2) narrator-disambiguation and biographical feature extraction (rijāl metadata), and (3) supervised and graph-based ML models to predict indicators relevant to traditional hadith classification (e.g., continuity of chain, possible breaks, ambiguous narrator identity). We review existing datasets (AR-Sanad, SanadSet, Multi-IsnadSet) and prior computational approaches — from heuristic graph-representations and HMM/POS methods to transformer-based models fine-tuned for Arabic and classical names — identifying strengths and gaps in coverage, annotation quality, and interpretability. Using a pilot corpus built from canonical collections (sampled variants from Ṣaḥīḥ collections and parallel manuscripts), we present experimental results that show automatic narrator-disambiguation and continuity-flagging can reach useful support-levels for human scholars (high recall in candidate linking; moderate precision in automatic grading), while fully automated final grading remains unreliable without expert oversight. We discuss key challenges: variant name forms and orthography, homonymous narrators across regions/periods, biased training data, and the normative complexity of mapping computational scores to juristic categories (ṣaḥīḥ, ḥasan, ḍaʿīf). The paper concludes with an ethical and methodological roadmap for integrating ML tools into hadith scholarship: transparent, explainable models; scholar-in-the-loop workflows; standardized datasets and annotation guidelines; and safeguards to avoid overreliance on automated judgments. The study demonstrates that ML is a promising assistive technology for isnād analysis but cannot replace human critical evaluation in hadith grading.