Glossary
Data annotation: giving data usable meaning
Annotation means giving data the label a model must learn to recover. It is a profession in its own right, and annotation quality caps model quality. It rests on a written guide and on a measure of agreement between annotators.
The label the model learns
Annotating means attaching to each observation the information a model must learn: a document's category, the useful area of an image, the intent of a sentence. The work rests on a guide that settles ambiguous cases in advance, and on a measure of agreement between annotators. The data and annotation operator qualification covers this role over 600 taught hours.
- A guide settles ambiguous cases before annotation begins
- Agreement between annotators is measured and corrected
- 600 taught hours in the matching professional qualification
What makes annotation reliable
The Artificial Intelligence programme has candidates write an annotation guide, annotate a shared batch across several people, then measure the agreement reached and revise the rules. The Artificial Intelligence and Business Decision certificate draws the steering consequence: how much annotated data for which expected quality level.
| Step | What gets decided there |
|---|---|
| The annotation guide | The categories retained and how ambiguous cases are handled |
| The calibration batch | One sample annotated by several people, to measure agreement |
| Production | The volume annotated, with a share double-checked |
| Revision | Rules adjusted when the same pattern of disagreement recurs |
The programmes and the qualification
2 catalogue programmes put “Data annotation” to work: 135 credits and 1350 taught hours in total. The official rule holds throughout — one credit stands for 25 hours of work, 10 of them taught. Every line below is recomputed from the programme page: level, duration, credits, taught volume and fees in Guinean francs appear exactly as filed in the official catalogue.
| Programme | Level | Duration | Volume | Fees |
|---|---|---|---|---|
| Artificial Intelligence programme — Bac+5 level | Master's degree | 2 years | 120 credits · 1200 taught hours | 7 800 000 GNF per year |
| Certificate in Artificial Intelligence for Business Decisions | Professional certification | 5 months | 15 credits · 150 taught hours | 3 600 000 GNF in total |
What we get asked about “Data annotation”
Does annotation require a technical background?
It requires rigour and an understanding of the domain being annotated. The matching professional qualification opens at that level.
How much data needs annotating?
The volume depends on task difficulty and the quality level targeted. The programme has you measure that relationship on a real case.
How do you keep annotators consistent?
Through a written guide, a shared calibration batch and an agreement measure repeated at regular intervals.
Explore next
- Machine learningMachine learning: principle, supervised learning, evaluation on held-out data and deployment, in IHETC programmes.
- EncryptionEncryption: symmetric and asymmetric, data in transit and at rest, key management, in IHETC cybersecurity and networking programmes.
- FirewallFirewall: filtering policy, segmentation, logging and periodic rule review, in IHETC security and networking programmes.
- Personal dataPersonal data: definition, purpose, minimisation, retention period and individual rights, in IHETC legal and security programmes.
- User experienceUser experience: field research, journeys, usability tests and iterations, in the IHETC Design and User Experience programme.
- MockupMockups and prototypes: fidelity levels, use in testing, handover to build, in IHETC design and digital creation programmes.