Developer utility for existing Tesseract output
Already have a Tesseract TSV? Inspect it locally.
This free tool turns Tesseract TSV output into a word-review queue: confidence distribution, page-level risk, weak tokens and a downloadable JSON report. It does not read or recover the original document.
Two separate workflows
Diagnostic utility, not a recovery preview.
For existing Tesseract users
Inspect Tesseract word confidence.
No TSV selected. The public sample is available for an instant test.
Interpretation boundary: confidence helps prioritize review. It is not ground-truth OCR accuracy and does not test reading order, table structure or field relationships.
Page summary
| Page | Words | Mean | Below 50 |
|---|
Lowest-confidence words
| Word | Location | Confidence |
|---|
Generate compatible TSV
Use Tesseract's TSV output.
The TSV config produces one row per hierarchy level and recognized word. Word rows carry the usable confidence values; structural rows commonly use negative confidence.
tesseract input.png stdout -l eng --psm 6 tsv > output.tsv
Beyond confidence
A high score can still hide a broken document.
The benchmark W-4 run reaches 90.04 mean confidence while plain text still cannot reliably preserve checkbox state or label-to-field relationships. For tables, forms and historical records, source-aware structural review remains necessary.