Researchtemporal annotationsclinical nlptransformerssynthetic data
Researchers Release TimeML-Compliant German Clinical Corpus
9.2
Modersohn et al. (J Med Internet Res, 2026) present a TimeML-conformant annotation schema and apply it to two German clinical corpora, producing 3000PAJ-temp (non-distributable) and GraSCCo-temp (public synthetic). They report high NER interannotator agreement (F1=0.90) and trained BERT-based baseline taggers achieving NER F1 between 0.64–0.85 and temporal relation F1 between 0.60–0.64, enabling temporal extraction for German clinical NLP.
Key Points
- 1Create TimeML-compliant annotations for German clinical text across real and synthetic corpora
- 2Achieve high NER interannotator agreement (F1=0.90), but lower temporal-relation agreement (F1≈0.41–0.57)
- 3Provide pretrained BERT baseline taggers attaining NER F1 0.64–0.85 and relations 0.60–0.64
Scoring Rationale
High novelty and practical utility from first TimeML German corpus; scope limited to German clinical domain.
Sources
Public references used for this report.
Practice with real Health & Insurance data
90 SQL & Python problems · 15 industry datasets
Used by DS/ML engineers at top companies
250 free problems · No credit card
See all Health & Insurance problems