The two sides
Printed OCR is the page-faithful transcription from editions/ee/mn/txt, with one line per printed line. Witness is MN01_Ee_Ee.09.txt, an independent Ee text used only as a cross-check. The scan is the only truth.
What gets skipped
Running heads (rend="head"), lone section numbers such as 3., and printed rules are displayed but excluded from the alignment. The witness carries no running heads.
Large witness-only blocks
The witness often expands peyyāla abbreviations that the printed page leaves as p., h. p., v. p. s. and similar. Those green blocks are edition-format differences, not necessarily transcription mistakes. Check them against the scan.
Alignment
Characters are aligned with a Myers shortest-edit-script diff, then edits are projected back onto each side’s native line breaks. A substitution is one deletion and one insertion at the same place. The approach follows the sanskrit-alignment idea of aligning the stream first and mapping to lines second.
Defaults
Case, punctuation, spacing and line-break hyphenation are ignored; diacritics are not. The witness drops punctuation and joins compounds, so this keeps the signal on letters. Change any of this under Comparison options.
Privacy
The gate and the comparison run entirely in this browser. No text is sent anywhere. The password gate is a convenience, not server-side access control.
Keys
j/k page · d diffs · s scan · [/] difference · / find · Esc close