An Ensemble Semantic Text Representation with Ontology and Query Expansion for Enhanced Indonesian Quranic Information Retrieval

This study explores the effectiveness of an ensemble method for Quranic text retrieval, aimed at improving the relevance and accuracy of verses retrieved for specific themes. The ensemble approach integrates three semantic models—Word2Vec, FastText, and GloVe—through a voting mechanism that considers verse frequency and semantic alignment with the query topics. Testing was conducted on themes such as prayer, zakat, fasting, umrah, and eschatology, reflecting fundamental aspects of Quranic teachings. Results demonstrate that the ensemble method significantly outperforms non-ensemble approaches, achieving an average relevance rate of 88%, compared to individual models (Word2Vec: 75%, FastText: 80%, GloVe: 82%). The ensemble method effectively combines the unique strengths of each model. Word2Vec captures general semantic relationships, FastText handles morphological nuances, and GloVe identifies global contextual patterns. By combining these capabilities, the ensemble approach improves both the quantity and quality of retrieved verses, making it a robust tool for semantic analysis in Quranic studies. This research contributes to the field of computational Islamic studies by demonstrating the practical advantages of ensemble methods for religious text retrieval. It lays the foundation for further advancements, including the integration of deep learning techniques, dynamic query handling, and cross-linguistic analysis. The ensemble method offers a promising framework for supporting more accurate and contextually relevant Quranic studies, promoting a deeper understanding of Islamic teachings through data-driven methodologies.

@article{Trisnawati2025,
  title     = {An Ensemble Semantic Text Representation with Ontology and Query Expansion for Enhanced Indonesian Quranic Information Retrieval},
  journal   = {International Journal of Advanced Computer Science and Applications},
  volume    = {16},
  number    = {1},
  year      = {2025},
  publisher = {The Science and Information Organization},
  author    = {Liza Trisnawati and Noor Azah Binti Samsudin and Shamsul Kamal Bin Ahmad Khalid and Ezak Fadzrin Bin Ahmad Shaubari and Sukri and Zul Indra},
  doi       = {10.14569/IJACSA.2025.0160148},
  url       = {https://doi.org/10.14569/IJACSA.2025.0160148}
}