A Graph-Theoretic and Semantic Embedding Approach to Religious Text Fact-Checking: Mathematical Frameworks for Hadith Verification

Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).

Fact-checking for religious text, specifically hadith, is difficult due to some rules in the text that often have exceptions and conflict evidence.Standard rule-based fact-checking systems are too rigid to handle these details, while pure machine learning models cannot clearly explain their decisions, which is required for hadith texts.To address this problem, this study develops a hybrid system for Hadith verification.This study focuses on food and drink chapters with English translation from the six authentic books of hadith (Sahih Bukhari, Sahih Muslim, Sunan Abu Dawood, Sunan Al-Tirmidhi, Sunan An-Nasai, Sunan Ibn Majah), using a test set of complex claims.The method involves three steps: first, we build a structured Knowledge Graph (KG) to clearly map out food items, prophet actions, and religious rulings.Second, we retrieve contextually relevant evidence using a pipeline that combines structural graph traversal with semantic embeddings.Finally, we classify the claims using machine learning models that are guided by the ontological rules in our KG.This proposed system achieves highest overall accuracy, 87.0% for Random Forest model.It also reaches 92.9% accuracy with conflicting evidence, performing much better than rule-based baseline and Support Vector Machine (SVM) algorithms.This study shows that combining crafted rules with flexible machine learning model creates a highly accurate and transparent system for verifying hadith texts.