Learn more about CommunicateIt!!!
I worked with Pragya Mishra on a research project for our Natural Language Processing class at Kennesaw State University.
Abstract :
This project develops and evaluates a question-answering (QA) system designed to address theological and interpretive questions about the New Testament. It uses a Retrieval-Augmented Generation (RAG) framework that integrates a pretrained large language model with a structured knowledge base consisting of public-domain Berean Standard Bible (BSB) New Testament and New Testament commentaries. User queries are embedded to retrieve semantically relevant passages, which are then supplied as contextual input for answer generation. The system is evaluated based on retrieval quality, answer faithfulness, and comparison to ground truth. Performance is benchmarked against a baseline BM25 keyword retrieval system without commentary, demonstrating that commentary-augmented semantic retrieval improves interpretive accuracy and reduces hallucinated responses.
This research was presented at the C-day conference at Kennesaw State University.
Introduction
Understanding the Bible presents a persistent challenge for new learners, particularly when interpreting the cultural, historical, and theological context behind individual verses. Many turn to generic large language models (LLMs) for assistance; however, these models are prone to hallucinated or misleading responses, especially when dealing with ancient texts. This limitation is amplified in biblical studies, where the nuances of Hebrew, Greek, and Greco-Roman contexts are not adequately captured in standard pretraining data, leading to incomplete or inaccurate interpretations.
This project addresses these challenges by developing a question-answering system that enhances Bible comprehension through structured contextual retrieval. The system breaks down verses and supplements them with relevant historical and theological insights drawn from curated sources. Both a baseline and an improved model are implemented and evaluated on a set of New Testament questions spanning factual, parabolic, and theological categories. In doing so, the project contributes to ongoing work in natural language processing, particularly in question answering over historical and religious texts.