AI Document Intelligence
OCR for Arabic and English, handwriting included, with automatic classification and data extraction.
Overview
InfoGlobe Document Intelligence turns scanned archives into searchable, structured data. It reads Arabic and English, printed and handwritten, including old and faded pages. Every document is sorted by type and the fields your teams need are filled in, all on servers inside your organization.
Accuracy comes from agreement, not from a single reading. Several models read each page independently, and a value is confirmed only when they agree; the rest goes to a reviewer who sees the page region behind it. In an independently graded sample, 93-98% of confirmed values were correct, and every correction stays in the audit trail.
The platform is already in production on one of the region's largest archiving projects, where it processes about 1,375 documents a day on a single GPU server. Results load into Docs2Keep or into the archive system you already use. Because it runs locally, sensitive government and personal records never leave your network.
What changes for you
An archive you can search
Staff find a record by name, date or reference number instead of opening files and boxes.
Checking instead of typing
Fields arrive pre-filled, so your team confirms values rather than typing them from scratch.
Data you can trace
Every value links back to the page region it came from, and every correction is on record.
What we deliver
-
Reads Arabic and English
Printed text and handwriting, including old and faded pages. Handwriting is rated conservatively, and its results depend on page quality and script.
-
Sorts by document type
We have deployed 16 types for government correspondence and about 30 for banking archives. Your list is agreed with you and tuned to your own records.
-
Extracts the fields you need
Date, subject, sender, receiver, organizations, people, ID and reference numbers. Hijri dates are converted and names are recorded in both languages.
-
Confirmed by agreement
Several models read each page independently. Only the values they agree on are confirmed automatically.
-
Review built in
Reviewers see the page region behind each value, with a confidence score for every field. Their corrections are kept and never overwritten.
-
Runs on your servers
Local AI vision models, with no cloud service required. Every view and change is written to a full audit trail.
How we work
Restore and scan
Fragile pages are restored first, then scanned at archive quality on your premises. If your archive is already scanned, we start from your images.
Read and classify
Local models read every page. Each one is then sorted by type, using the list agreed for your archive.
Extract and confirm
Fields are extracted from each document. A value is confirmed when the models agree on it.
Review and archive
Reviewers check the remaining values against the page. Everything lands in your archive with its audit trail.
Selected work
-
Building files and engineering drawings for the municipal e-archive, delivered in two phases.
Questions
Can it read Arabic handwriting?
Yes, in Arabic and English, although handwriting is harder than print. We rate it conservatively at 65-79%, depending on page quality and the age and style of the script. Fields are pre-filled and a reviewer confirms them.
What do we need to prepare?
A representative sample of your archive, the document types you hold and the fields your teams need. During the assessment we agree the scope with you and size the servers the models will run on.
Where does our data stay?
On your own servers or in your data centre. The models run locally and no cloud service is required, so pages and extracted data stay inside your network.
Other solutions
- Secure Localized AIAI that runs inside your network to classify sensitive documents, spot unusual access and help prevent data and document leaks.
- AI AgentsAgents that handle routine work end to end, with a person approving the steps that matter.
- Digitization, Restoration and e-ArchivingRestoration of fragile and historic documents, then on-site scanning and indexing for archives of any size.
- VR and Immersive ExperiencesTraining simulations, virtual tours and complete virtual worlds such as MetaNileVerse.
- Software, Portals and MobileWeb, cloud and mobile applications built around your workflows.
- Integration and Data MigrationConnect legacy systems and move data without losing a record.
- ConsultingDigital infrastructure and organizational development, from strategy to rollout.
- Branding and UXVisual identities and interfaces people find easy to use.
Plan your project with us
Tell us what you have and where you want to be. We will propose a scope, a timeline and a pilot.