01
Language Representation
Underrepresented and low-resource languages receive uneven support from mainstream AI and OCR systems.
NASHIR
Making Underrepresented Media Archives Accessible, Searchable and Verifiable with AI
An AI-powered document intelligence and media archive platform focused on underrepresented languages, historical documents, and evidence-based access to information.
In development for over two years, built and tested through real-world work in Iraq, with a focus on challenging historical documents and underrepresented languages.
AI should not replace the source.
It should help people reach it.
Built as a working system, not a concept.
Figures below describe results on selected internal tests and the scope of the current collection. They are not universal benchmark claims.
400
pages processed in a single batch
5
languages tested
Up to 91%
on selected printed-document recognition tests
Up to 83%
on selected handwriting recognition tests
~150 years
of Iraqi press history
Thousands
of newspapers and publications
Live Demo
Explore the Live Platforms
Al-Nasikh and the Iraqi Press Encyclopedia are working systems, not concepts. Reviewers can try both directly.
Access the live demo
Demo credentials are provided for reviewers. The account is read/demo-safe, has limited permissions, and contains no personal or confidential data.
- Username
- demo.reviewer@uipro.me
- Password
- Nashir-Demo2026
The Problem
The Digital World Does Not Read Every Language Equally
Digital transformation has not reached every language, script, and historical archive equally. Modern AI and OCR systems often perform best where large amounts of digital training data and commercial attention already exist.
Historical newspapers, scanned books, lithographic prints, degraded documents, and handwritten records introduce additional challenges.
For many communities, the problem is not that their history does not exist. The problem is that their history remains difficult for machines — and therefore difficult for modern search, analysis, and AI systems — to access.
When historical information cannot be reliably digitized, entire communities face barriers to searching, studying, understanding, and verifying their own media history.
02
Historical Documents
Historical printing methods, degraded pages, unusual layouts, and handwriting create additional recognition challenges.
03
Access & Verification
Making information searchable is not enough. Users also need a way to return to the original source and verify what they see.
The Solution
One Mission. Two Connected Layers.
Al-Nasikh provides the document intelligence layer. The Iraqi Press Encyclopedia demonstrates what becomes possible when historical media can be digitized, searched, explored, and connected back to its original evidence.
Layer 01 — Technology
Al-Nasikh
AI-powered document intelligence for underrepresented languages and challenging historical documents.
Layer 02 — Real-World Application
Iraqi Press Encyclopedia
A searchable digital implementation of a three-volume Iraqi press encyclopedia covering roughly 150 years of press history.
Layer 03 — Verification
Evidence
Original pages, publication information, extracted text, and source context remain accessible to the user.
01Historical Documents
02Al-Nasikh
AI document recognition
03Digital Text
04Iraqi Press Encyclopedia
05AI-Assisted Discovery
06Original Source
07Verification
Technology Layer
Al-Nasikh
AI document intelligence for languages and documents that are harder for mainstream systems to read.
Al-Nasikh is a working online platform developed to improve text extraction from underrepresented languages and challenging document types.
The platform is currently Arabic-first and has been tested across Arabic, Hindi, Hebrew, Swahili, and Javanese.
It is designed not only for modern printed documents, but also for historical printed material and Arabic handwriting.
- Historical and lithographic printed documents
- Arabic handwriting recognition
- Support and experimentation across multiple underrepresented languages
- Document preprocessing and recognition pipeline
- Adaptation of open-source models
- Custom improvements to document recognition and processing
Technical noteAl-Nasikh builds on open-source models and extends them through custom document-processing, preprocessing, recognition, and language-focused methods. The system is designed around the needs of underrepresented languages and challenging document collections.
Evidence from testing — selected test results
Up to 91%
Selected printed-document recognition tests
Up to 83%
Selected handwriting recognition tests
Results vary by document quality, script, layout, and test set.
Process entire books, not just individual pages.
Up to 400 pages in a single batch.
Arabic Handwriting
Tested on good-quality handwritten Arabic documents.
Historical Printed Documents
Tested on historical printed and lithographic documents.
Handwritten Documents
Tested on selected good-quality handwritten Arabic documents.
Languages Tested
- Arabic
- Hindi
- Hebrew
- Swahili
- Javanese
Before / After: document to text
Example 1 — Historical Printed Document
Example 2 — Arabic Handwriting
Real-World Application
Iraqi Press Encyclopedia
A real-world application of the technology.
An original three-volume encyclopedia documenting Iraq's press history has been transformed into a searchable digital knowledge platform.
The collection covers roughly 150 years of Iraqi press history and includes thousands of newspapers and publications.
Search
Search across digitized press content using words and text.
Date
Explore publications by publication date and historical period.
Publication Type
Filter and explore different types of publications.
Classification
Explore the collection through available classifications.
People & Entities
Explore information associated with people and other entities where available.
Original Source
Open the original encyclopedia/document page as an image.
AI-Assisted Exploration
Use AI to help explore and understand the collection while maintaining a connection to source material.
Search to Source
From a Question to the Original Source
A real search in the Iraqi Press Encyclopedia stays connected to its original page at every step.
01User Question
02AI-Assisted Search
03Result
04Publication, Date, Page
05Original Page Image
06Extracted Text
07Verification
Illustrative view of the encyclopedia's homepage alongside an archival source page. Explore the live platform for a complete search-to-source walkthrough.
Evidence
From AI Answers to Verifiable Evidence
We do not want users to simply trust an AI-generated answer. We want AI to help them reach the evidence.
Every useful result should remain connected to the material from which it was derived.
See the Source Behind the Information
Instead of presenting information as an isolated AI response, the platform can lead users back to the original page, publication, date, and extracted text.
Evidence Panel
- Publication
- Name of the newspaper or publication
- Publication date
- Date of the original issue
- Page number
- Page reference within the source
- Extracted original text
- Digitized text from the page
- OCR confidence
- Recognition confidence score
- Related sources
- Linked references, where available
01Question
02AI-Assisted Search
03Result
04Source
05Original Page
06Verification
AI → Answer → Evidence → Verification
Media & Information Literacy
Designed for Media and Information Literacy
Media and Information Literacy is not only about accessing information. It is also about understanding where information comes from, evaluating it critically, and making informed judgments about what to trust and share.
Our approach uses AI as an access and discovery layer while preserving the user's ability to inspect the underlying evidence.
The MIL layer is not an additional feature. It is built into how information is presented: search → context → source → verification.
Find
Discover information that was previously difficult to search.
Understand
Use AI-assisted tools to explore large historical collections.
Question
Examine context, metadata, sources, and uncertainty.
Verify
Return to the original document before accepting or sharing information.
AI should expand access to evidence — not replace evidence.
Impact
Why It Matters
The project connects technical innovation with a public information need: making historical media more accessible without losing the connection to the original record.
01
Journalists
Access historical reporting and primary media sources that are difficult to search digitally.
02
Researchers
Search and analyze historical media collections at a scale that would be difficult through physical archives alone.
03
Students & Young People
Develop stronger habits of source-based research and critical engagement with information.
04
Underserved Language Communities
Help bring languages and documents with limited AI representation into the digital information ecosystem.
05
Archivists & Cultural Institutions
Help cultural and media archives become more accessible, searchable, and useful to researchers and communities.
From Iraq to Other Underrepresented Archives
Iraq is the starting point — not the boundary.
The same approach can be adapted to other historical newspaper collections, cultural archives, and underrepresented languages around the world.
Potential future applications — not yet supported
- Historical newspapers
- Local archives
- Cultural heritage collections
- Minority and low-resource languages
- Academic collections
- Community archives
Approach
What We Focus On
Conventional AI/OCR Challenges
- Uneven language support
- Modern-document focus
- Limited support for difficult historical documents
- Information can become detached from original context
Our Focus
- Underrepresented languages
- Historical documents
- Large batch processing
- Arabic handwriting
- Searchable media archives
- Original source visibility
- Evidence-connected AI exploration
Future Direction
From One Archive to a Global Network of Accessible Media Memory
Our long-term vision is to help cultural and media archives become machine-readable, searchable, and accessible while preserving the connection to their original sources.
Starting with Iraq, the technology can evolve toward broader support for underrepresented languages, historical publications, and community archives.
More Languages
Expand support for additional underrepresented and low-resource languages.
More Archives
Apply the workflow to historical newspapers and cultural collections.
Better Verification
Strengthen source linking, provenance, and evidence-aware AI interactions.
These directions describe planned work. They are not presented as completed capabilities.
Project Status
Built. Tested. Now Ready to Scale.
Research
Completed
Prototype
Completed
Working Platform
Completed
Real-World Application
Completed
Scaling & Expansion
In progress
Al-Nasikh is already available as a working online platform.
The Iraqi Press Encyclopedia is an active digital implementation demonstrating how the technology can be applied to a large historical media collection.
The next stage is to expand the technology, improve language coverage, and make the approach applicable to more archives and communities.
Project Origin
Built from a Real Problem
This project grew from a practical challenge: how can historical Arabic and other underrepresented-language documents become accessible to modern digital systems without losing the evidence contained in the original documents?
The work has been under development for over two years, combining software engineering, document processing, OCR experimentation, digital archiving, and AI-assisted information retrieval.
Mahdi Ali Noori Alshammari
Project Lead
Software engineer leading the design and development of Al-Nasikh and the Iraqi Press Encyclopedia.
Ruya Ridha Baqer Aboaldigaj
Assistant
Software developer supporting the project's development and testing.
Hassan Ali Noori Alshammari
Archive & Product Development
Works on archive development and product direction for the Iraqi Press Encyclopedia.
UNESCO Youth Hackathon 2026
Play Your Part: Youth Designing the Future of Media and Information Literacy
NASHIR has been in development for over two years. In 2026, the project is taking part in the UNESCO Youth Hackathon 2026, at the intersection of Artificial Intelligence and Media and Information Literacy.
Call to Action
Help Make More of the World's Information Accessible
The technologies that shape our information environment should not leave languages, communities, or historical records behind.
Al-Nasikh begins with document access. The Iraqi Press Encyclopedia demonstrates what becomes possible. The larger goal is a more inclusive and verifiable information ecosystem.