An Evidence-Based Institutional Pipeline for Source-Grounded Question Answering over Islamic Religious Education Books

Abstrak

This preprint presents and evaluates an institutional pipeline for transforming born-digital Islamic Religious Education books in Portable Document Format into indexed knowledge units for source-grounded question answering. The implementation integrates a digital library, metadata management, an artificial intelligence knowledge bridge, a retrieval-augmented generation service, an artificial intelligence question-and-answer module, and a website chatbot. The study applies a single-case, artifact- and evidence-based design. Three documents representing standard text, Indonesian-Arabic content, and complex visual layout were evaluated using 20 System Integration Testing scenarios, 10 role-based User Acceptance Testing scenarios, and 10 source-to-answer checks. The results are limited to the frozen test build, three books, and selected questions. They should not be interpreted as overall chatbot accuracy, user satisfaction, or learning effectiveness. This record contains an author manuscript/preprint and is not the version of record.

Identifier

• DOI: https://doi.org/10.5281/zenodo.21754219
• OpenAlex: https://openalex.org/W7172218194