TLDR: SpiritRAG is a novel AI-powered Q&A system designed to simplify research on religion and spirituality (R/S) within the United Nations archive. Utilizing Retrieval-Augmented Generation (RAG), it processes approximately 7,500 UN resolution documents on health and education to provide accurate, context-sensitive answers through an interactive, chat-based interface. Evaluated positively by domain experts, SpiritRAG aims to make complex archival information more accessible for researchers and policymakers, with future plans to expand its document sources.
Researchers and policymakers often face significant challenges when trying to extract information on complex and context-dependent concepts like religion and spirituality (R/S) from vast archival sources. Traditional search methods can be time-consuming and require specialized knowledge, especially when dealing with large, noisy datasets like those found in the United Nations (UN) archive.
To address this, a new interactive Question Answering (Q&A) system called SpiritRAG has been developed. SpiritRAG is designed to make complex, context-sensitive database searches much easier, providing an accessible, chat-based web interface for users.
What is SpiritRAG?
SpiritRAG is built using Retrieval-Augmented Generation (RAG), a technique that combines information retrieval with advanced language model capabilities. This allows the system to not only find relevant documents but also to generate coherent and contextually sensitive answers based on the retrieved information. The system was developed using approximately 7,500 UN resolution documents related to R/S in the domains of health and education, available in seven languages including Arabic, Chinese, English, French, German, Russian, and Spanish.
The core of SpiritRAG involves two main components: document retrieval and answer generation. When a user asks a question, the system first identifies and retrieves the most relevant UN resolution documents from its extensive corpus. It then uses a powerful language model, specifically instruction-tuned Qwen3 models, to synthesize information from these retrieved documents and formulate a comprehensive answer. A unique feature of SpiritRAG is its ability to incorporate user-provided PDF documents as additional source material, allowing for even more tailored responses.
Ease of Use and Deployment
SpiritRAG is designed to be lightweight and easy to deploy, whether locally or on cloud services, requiring minimal computational resources. Its user-friendly interface allows for real-time human-machine interactions in a chatbot setting. Users can ask questions related to religion and spirituality, and the system provides answers rendered in Markdown for improved readability. Additionally, users can view the most relevant source documents used for answer generation, complete with metadata like document title, key subjects, publication date, and available languages. Clicking on an available language opens a PDF viewer to display the original UN resolution document.
Also Read:
- Uni-RAG: A New AI Framework for Multi-Modal Learning in STEM Education
- Advancing RAG Through Progressive Instruction Tuning
Evaluation and Impact
A pilot test and evaluation of SpiritRAG were conducted with domain experts specializing in R/S and education research. The experts assessed the system’s document retrieval and answer generation capabilities across various dimensions, including relevance, accuracy, usefulness, and coherence. The results were promising, with evaluators noting significant improvements in performance as larger language models were integrated. SpiritRAG was perceived as a highly innovative and potentially groundbreaking interactive knowledge engine, highlighting its potential for significant impact in public policy, historical scholarship, and digital humanities.
While SpiritRAG currently focuses on health and R/S-related questions within UN resolution documents, its modular architecture allows for future expansion to incorporate a broader range of document sources, including religious texts. This will further enhance accessibility to embedded spiritual knowledge and its relevance to contemporary discourse. For more details, you can refer to the full research paper: SpiritRAG: A Q&A System for Religion and Spirituality in the United Nations Archive.
