EXCLUSIVE UPDATE ON AI TRANSCRIPTION PROJECT

The CSNTM Indexing Project is moving steadily toward the development of an AI-powered tool that can identify the range of biblical text visible in manuscript images. In practical terms, the goal is to help determine which passage of Scripture appears on each page of a manuscript, making these images easier to search, study, and use.

Our initial testing has been very encouraging. We began by asking several major AI models to identify the biblical passages visible on a small set of test images. Some models performed better than others, but the strongest results were impressive: the best models correctly identified the passage with over 90% accuracy. These early results suggest that, with careful preparation and well-designed instructions, AI can become a powerful aid for manuscript indexing, dramatically speeding up work that would otherwise take many hours.

The project’s current focus is building a larger, carefully verified sample set. CSNTM already has human-generated indexing data for more than 1,000 manuscript images. This data is now being checked to confirm that the listed Scripture ranges truly match the text visible in each image. Though time consuming, this step is essential because any AI tool must be evaluated against reliable answers before we can know how well it is performing. Once verified, this sample set will serve as a sort of “answer key”, allowing us to measure the tool’s accuracy and identify where it still needs improvement.

The potential impact of this work is truly exciting. If successful, this tool could help scholars locate biblical passages in manuscript images more efficiently, while also making CSNTM’s digital collections more accessible to pastors, students, and anyone who wants to explore the manuscript history of the New Testament.

I look forward to sharing more updates in the months ahead.

With gratitude,