Data labelling
Marking up images, text and audio so a model can learn from it. Repetitive by nature, and quality matters more than speed because a mislabelled example teaches the wrong thing.
What it involves
- Drawing boxes, tagging categories, or transcribing short clips.
- A guideline document with edge cases, which is worth reading properly.
- Batches, usually with a qualification round first.
What gets rejected
Stated up front, because a rejection reason you only discover afterwards is not a standard, it is a trap.
- Disagreeing with the gold-standard items mixed into the batch.
- Inconsistency between similar examples.
- Skipping the edge cases the guidelines specifically cover.
What it pays
Priced per item, so your effective rate depends on how quickly you can be accurate. It rises sharply once you know a project's guidelines, which is why we surface repeat batches from projects you have already qualified on.
We do not publish an hourly figure for this or any other category. The honest answer depends on your country, your languages and what clients are asking for this week, and a number on a marketing page would be the best case dressed up as the typical one.