Project
Saheeh — Arabic Manuscript Pipeline
Turns scanned Arabic manuscripts into clean, structured, verifiable data
Go
SQLite
Tesseract
SP
Overview
About this project
A Go command-line pipeline that turns scanned Arabic Islamic manuscripts into structured, trustworthy text. OCR output for classical Arabic is notoriously messy — broken diacritics, split words, confused letter forms — so Saheeh treats cleaning as a graded, auditable process rather than a single pass.
The full pipeline runs ocr → ingest → clean → score → repair → export. Every document gets a confidence score, so you can see which passages are reliable and which still need a human eye, and a directory-wide scorecard reports the state of an entire corpus at a glance.
Written in Go with no CGO and a single pure-Go SQLite dependency, so it builds and runs anywhere.