fusus's Introduction

Fusus

This is a workflow that transforms scanned pages into readable text.

The pages come from several printed Arabic books from the past few centuries.

The workflow takes care of cleaning, OCR, proofing, converting to tab separated files and from there to Text-Fabric from where the text material can be processed further.

Features

cleaning is included: specks and symbols can be specified for cleaning by copying and pasting such fragments and storing them in a designated directory;
column layout and line boundaries are detected prior to OCRing;
individual lines will be passed to the OCR engine, which is Kraken using a model trained on many printed Arabic books, see model;
the results are stored in tab-separated files, retaining boundary boxes and confidences;
proofing pages can be generated for manually checking the OCR results;
the OCR results of each book are composed into Text-Fabric datasets.

This lays the foundations for:

correcting OCR mistakes;
enriching the text with morphological/linguistic annotations, named entities;
perform intertextuality research between the ground work (the "Fusus" by Ibn Arabi) and its commentary books.

A lot of cleaning has been carried out on two editions of the Fusus: Lakhnawi and Afifi. After that these editions have been aligned and brought together in a single dataset, in which it is possible read back the individual editions.

Text-Fabric interface

Get started with the tutorial.

We also have generated a static search interface.

Just click fusus-search and off you go.

You can do full text search via regular expressions, not only in the full-text, but also in attributes of the text, notably the bounding box information of each word.