<![CDATA[This research was conducted with the intention of developing an English-language hadith search system that is not only syntactically accurate, but also contextually appropriate. The system was developed using a combination of convolutional neural networks (CNN) and two text representation methods, namely Term Frequency–Inverse Document Frequency (TF-IDF) and Sentence-BERT (SBERT). CNN is used to classify hadiths into seven main categories based on chapter titles. In the semantic retrieval stage, TF-IDF and SBERT were utilized to represent the text of the hadith and user queries, then both were evaluated using cosine similarity. Testing was conducted using five queries commonly used in Islamic studies, then evaluated manually for semantic similarity. As a result, the tuned CNN achieved a classification accuracy of 94%. On the other hand, although the TF-IDF approach produced greater similarity results, SBERT proved to be superior in generating more relevant results in semantic searches. These results indicate that TF-IDF is superior in terms of speed, but SBERT is better at understanding sentence context in depth. This research contributes to the development of a meaning-based hadith search system and emphasizes the importance of a semantic approach in religious text search. Moving forward, system development can be directed toward multilingual support and evaluation on a larger scale.]]>
Copyrights © 2025
