Building Arabic food logging into a health app.

Why Arabic food logging is different

• Many everyday meals are mixed dishes, such as rice dishes, stews and stuffed vegetables, so a single database entry rarely fits.
• Shared platters are common, so the plate in the photo is often not what one person ate.
• Oil, ghee and butter go into the pot, not onto the plate, so a photo hides much of the fat.
• The same dish changes from one country and one family to the next.
• People describe food in their own dialect, and spell dish names in different ways in Arabic and in Latin letters.

Get the right-to-left interface right

• Set the direction on the content (dir="rtl" and lang="ar"), and use logical CSS properties such as margin-inline-start so the layout mirrors without a second stylesheet.
• Mirror directional icons such as arrows, chevrons and progress steps. Leave icons that are not directional, such as a camera or a microphone, as they are.
• Isolate Latin text inside Arabic sentences, such as a brand name or a unit, so the order of words does not jump. The HTML bdi element or Unicode isolates handle this.
• Choose one style of digits, Western or Arabic-Indic, and use it across the app, including charts.
• Test with real Arabic copy, not translated placeholder text, and with long dish names.

Accept Arabic text and voice

Typing a dish name into a search box is where many users give up. Let people describe a meal the way they would say it, in Arabic, by text or voice, and let the service work out the foods.

The BiteRight API accepts Arabic for photo, voice and text logging. Voice logging sends a short audio note with the meal date, and needs the user's dietary profile set up first.

How voice food logging works →

Handle shared dishes and hidden fats

• For a shared platter, ask how much of it the person ate, with simple options such as a quarter or a half, or a weight in grams.
• Prompt for cooking fat on mixed dishes: oil, ghee, butter or cream added during cooking.
• Let users save a family recipe once, so a dish they eat every week is quick to log the same way.
• Show results as estimates, because the same dish can vary a lot between kitchens.

Use regional food data

Nutrient values for a regional dish are only as good as the food data behind them. Ask any provider which food-composition databases it uses, and whether it covers the countries your users live in.

BiteRight's food data includes a Bahraini food-composition database and localised Middle Eastern dishes, alongside national databases from the US, UK, France, Australia and Canada. Mixed dishes are broken into likely ingredients before nutrients are looked up.

Arabic food recognition at BiteRight →

Test with the meals your users eat

Recognition accuracy for Arabic and Middle Eastern food has not been published by BiteRight, and published figures from any vendor rarely match your users' meals. Build a test set of typical home and restaurant meals from your market, weigh the reference portions, and compare services on the same set.

A test plan for food recognition APIs →