Service
Annotation & labeling
Labels designed with your ML team and applied by Khmer specialists to text, audio and images, with agreement measured on every batch.
What's included
- Intent, entity and sentiment labelling
- Khmer word segmentation and part-of-speech tags
- Audio event and emotion tags
- Bounding boxes, polygons and key-points
- Label-guide design and edge-case handling
Sample deliverable
Intent and entity labels on a banking request.
Illustrative example. Real samples come with your free pilot.
{
"text": "ខ្ញុំចង់ផ្ទេរប្រាក់ ៥០ ដុល្លារ ទៅ សុខា នៅថ្ងៃស្អែក",
"intent": "transfer_money",
"entities": [
{ "type": "AMOUNT", "text": "៥០ ដុល្លារ" },
{ "type": "PERSON", "text": "សុខា" },
{ "type": "DATE", "text": "ថ្ងៃស្អែក" }
],
"sentiment": "neutral", "reviewed_by": 2
}How it works
- 1
Guidelines
We write or adapt your label guide with Khmer examples.
- 2
Calibrate
Annotators label a test set until agreement reaches the target.
- 3
Label
Production labelling with hidden gold questions mixed in.
- 4
Review
A second-pass review and an agreement report for every batch.
Frequently asked questions
Can you use our annotation tool?
Yes. We work in Label Studio, CVAT, Labelbox or your in-house tool, or in our own secure set-up.
How do you measure quality?
With inter-annotator agreement (Cohen's κ) plus hidden gold questions. Both are reported with every batch.
Do you label languages other than Khmer?
Khmer is our focus. We also handle mixed Khmer–English text and, on request, minority languages spoken in Cambodia.
Related services
Start with a free pilot
We produce a small batch to your spec so you can check the quality before you commit.