Arabic quality review for software and AI products
I review software and AI products the way Arabic users see them: interface, AI answers, and right-to-left layout, and I tell product teams what to fix first.
What I check
Arabic interface text
Menus, buttons, error messages, notifications, and help content.
AI-generated Arabic answers
Accuracy, meaning, terminology, tone, and cultural fit, including confident wrong answers and misread dialect.
Right-to-left (RTL) layout
Alignment, mixed Arabic-English text, numbers, and dates.
Products I review
AI assistants and chatbots
Arabic answers from general and specialized AI models: meaning, accuracy, dialect, and tone. Based on my last three years of Arabic AI quality work at Oracle.
Business and cloud software
Arabic interfaces, messages, and help content in business applications. Based on 28 years of Arabic localization quality at Oracle, across 700+ products.
Voice, text-to-speech, and video dubbing
Arabic pronunciation, naturalness, and subtitles. Based on Arabic text-to-speech evaluation for Oracle Cloud (2025) and 100+ subtitled videos.
Payments and financial services
Numbers, currencies, dates, and the Arabic text on payment screens, where one error costs trust. Based on locale quality checks at Oracle: numerals, dates, and currencies in Arabic interfaces.
What you get
A prioritized list of findings. For each one:
- where it appears and how to reproduce it
- its severity, and why
- the effect on your users
- a recommended fix direction your team can act on
Severity follows impact, not appearance. The same error ranks higher on a payment button than in a footer.
How a review works
- Free check: tell me which product. I test it myself and send you two findings, free of charge.
- Short call: we go through the findings and agree which Arabic flows matter most to you.
- Diagnostic: a fixed-scope review, delivered in 5 business days, with every finding prioritized by severity.
How I work
I test the product myself, as an Arabic user would, and I report only what I observe and can reproduce.
I classify errors with MQM, the established error taxonomy and severity scheme. For Arabic language errors I also draw on LQM, a 2026 academic framework (Magdy et al., Findings of ACL 2026) that classifies Arabic errors across six linguistic levels. The right-to-left and bidirectional rendering checks are my own method.
Why me
At Oracle, I managed the quality of external vendor teams through style guides, terminology databases, and written feedback.
In my later years there, I applied the same quality methods to Arabic AI output: evaluating LLM responses for accuracy, terminology, and hallucination-related errors, and helping set validation standards for Arabic enterprise AI.
That experience shows me how Arabic text behaves once it enters a real software product, whether a translator or a model wrote it, so I judge it as users will meet it: in the interface, in context, and at scale.
I’m independent: when I review your product, I don’t sell you the fix, so my verdict has no conflict of interest.
Background
- 33 years in Arabic localization, 28 of them at Oracle, heading Arabic localization quality across 700+ products
- Arabic AI quality for the last 3 years, including Arabic text-to-speech evaluation for Oracle Cloud (2025)
- PhD in Library and Information Science (Arabic digital content management systems), Minufiya University
- Author/translator of 7 published technology and language books
- EN 15038-certified (2014; predecessor to ISO 17100)
Start with one product
Tell me which product, and which Arabic flows matter most to you. I’ll reply with how I’d review it.
Email info@dareenlingo.com, or message me on LinkedIn or WhatsApp.