Core responsibilities
- Build or qualify an annotated corpus
- Train a lexical or neural baseline
- Measure errors by language and document type
Technical skills
- Tokenisation and text representations
- Annotation guidelines and inter-annotator agreement
- Evaluating NER, classification or retrieval
Behavioural skills
- Bring annotators to agreement
- Explain common errors to domain experts
Typical deliverable
An entity extractor tested on multilingual documents and degraded input formats.
Possible assessment methods
Compare two methods on annotated text and analyse false positives with a plan to address them.
