Annotation types
Labels matched to your model, not a fixed menu.
Transcription
Verbatim or clean-read, in Arabic script, following your conventions.
Timestamps
Segment- or word-level timing aligned to the audio.
Speaker diarization
Who spoke when, across multi-speaker recordings.
Dialect labeling
Tag the dialect actually spoken, not just “Arabic”.
Intent labeling
What the speaker is trying to do, for voice agents and IVR.
Entity labeling
Names, places, numbers, products, and domain terms.
Emotion & tone
Where appropriate for the use case and clearly defined.
Code-switching
Mark where speakers move between Arabic and English.
Metadata enrichment
Add speaker, environment, and quality fields to existing audio.
Why native annotators
Arabic-native, not generic outsourcing.
A transcriber who doesn’t speak the dialect guesses at it. Our annotators are matched to the dialect of the audio, work to written guidelines, and are reviewed by someone else before anything is delivered.
-
01
Guidelines
We agree conventions for spelling, tags, and edge cases with you.
-
02
Annotation
Dialect-matched annotators work through the audio.
-
03
Review
A separate reviewer approves or returns each item with notes.
-
04
Delivery
Labels in your schema, with a QA summary.
Send us a sample of your audio.
We’ll annotate a pilot batch to your guidelines so you can judge the quality directly.