<![CDATA[Conventional Qur'anic exegesis retrieval systems that rely on keyword matching often return less relevant results because they are unable to capture the semantic meaning and context of users' queries. This study aims to develop a semantic search system for Tafsir Al-Misbah using the IndoBERT model within the CRISP-DM framework. The dataset consists of 6,236 Indonesian-language tafsir summaries processed through data cleaning, case folding, normalization, and tokenization. The IndoBERT model was then fine-tuned to generate semantic representations of user queries and document contents, while FAISS was employed to accelerate retrieval using cosine similarity. System performance was evaluated using Mean Average Precision (mAP), Mean Reciprocal Rank (MRR), and Normalized Discounted Cumulative Gain (nDCG). The evaluation results achieved Top-10 scores of 0.2128 for mAP, 0.2681 for MRR, and 0.2792 for nDCG, while the Top-50 scores reached 0.2301, 0.2803, and 0.3481, respectively. These results indicate adequate retrieval performance in the domain of Qur'anic exegesis and demonstrate that the proposed semantic search approach based on IndoBERT and FAISS is capable of providing contextually relevant retrieval results for users' queries.]]>
Copyrights © 2026
