In research
Fluent in the language of the nation.
Models that understand Arabic names, documents and dialects as natively as English.
The problem
Most AI is built in English first and adapted to Arabic later. The result shows in practice: Arabic names transliterated many ways, scanned Arabic documents read poorly, and dialects misunderstood.
Questions we are researching
- How can a system match Arabic names across their many English spellings without missing true matches?
- How accurately can scanned Arabic forms, contracts, identity documents and handwriting be read and checked?
- How well do models understand Gulf dialects alongside Modern Standard Arabic?
What we plan to build
Planned outputs. Each passes the lab's stage gates before it is offered to anyone.
- Name-matching research
- Methods to reduce false matches from transliteration, tested on realistic data.
- Arabic document intelligence
- Reading, extracting and checking Arabic and English documents.