Despite advances in multilingual information retrieval (MLIR), a gap exists between research and practical application. This study leverages the characteristics of the multilingual Quran corpus to investigate optimal strategies for developing an ad hoc IR system that can meet the specific user needs of the Islamic domain in multiple languages. Eleven retrieval models were prepared using four training approaches: monolingual, cross-lingual, translate-train-all, and a hybrid approach combining cross-lingual and monolingual techniques. Evaluation results on an internal dataset demonstrate that the hybrid approach achieves promising results across a variety of retrieval scenarios. Furthermore, we provide a detailed analysis of the impact of various training configurations on the embedding space and multilingual retrieval effectiveness. Finally, we discuss deployment considerations, highlighting the cost-effectiveness of deploying a single, lightweight, multipurpose model for practical MLIR applications.
