A Documentary Restoration of Scrivener’s 1887 Greek New Testament
Restored through repeated human collation, page-by-page image preparation, independent AI-assisted OCR, controlled difference detection, image-based adjudication, and a completed final audit of the Greek text.
Every textual feature included in the final audit was verified against the printed page, and every detected discrepancy was individually adjudicated before the final text was approved.
The Restoration Journey
The restored text was not produced by a single transcription or a single software pass. It emerged from a sequence of independent comparisons, documented decisions, controlled corrections, and complete rechecks.
A commonly circulated electronic transcription of Scrivener’s text served as the initial working copy. It was useful as a starting point, but its exact correspondence with the 1887 printed edition had not been independently established.
More than ten reviewers repeatedly compared the printed TR1887 source and the working text from Matthew through Revelation.
The textual area of every PDF page was manually cropped and preserved as a page image suitable for controlled OCR and visual verification.
AI-assisted OCR generated a new transcription from the page images, functioning as an independent witness rather than an automatic correction.
The system isolated differences in word sequence, spelling, capitalization, omission, addition, and ordering.
Each detected discrepancy was compared directly with the printed page and classified before any correction could enter the working text.
After orthographic agreement was established, the remaining textual features included in the final audit were separately examined and corrected.
Approved corrections were applied through a controlled workflow, and the completed text was rechecked against the original printed source.
Documentary Evidence
These archival photographs preserve evidence from the manual and digital collation process. Select an image to examine it at a larger scale.
Each archival image forms part of the restoration record and may be documented with its location, source page, discrepancy type, decision, and final verification record.
The complete, searchable audit trail — every discrepancy, source-image reference, and grapheme-level decision behind this restoration.
Methodology
The software located possible discrepancies and preserved their context. The printed image determined the reading, and only approved corrections were applied.
Page images were transcribed independently to create a second dataset capable of exposing differences overlooked during visual proofreading.
Separate passes isolated orthographic and surface-form questions so that one type of correction did not unintentionally alter another feature.
Candidate corrections remained pending until a reviewer examined the relevant printed page and explicitly approved, rejected, or deferred them.
After approved corrections were applied, the complete text was processed again to confirm that the final audit closed without unresolved discrepancies.
This is a simplified illustration. Inspect the real, page-by-page audit register →
The Completed Edition
The completed edition joins documentary accuracy with professional book design. Its Greek text, marginal references, textual notes, typography, and two-color visual system have been developed as a unified scholarly publication.