17,600 professionally annotated veterinary electronic health records (EHRs) from UK first-opinion practices are partitioned into training, evaluation, and test sets. The dataset features free-text clinical narratives mapped to International Classification of Disease 11 (ICD-11) syndromic categories.
Use Cases
- Train classification models to map free-text clinical notes to ICD-11 syndromic labels
- Benchmark the cross-clinic generalizability of NLP models using the partitioned evaluation and test sets
- Develop automated surveillance tools for identifying disease trends in small animal populations based on EHR narratives
Strengths
- 17,600 EHR records split into 11,000 training, 1,600 evaluation, and 5,000 test samples
- Professional annotations using the International Classification of Disease 11 (ICD-11) syndromic framework
- Free-text clinical notes sourced from diverse first-opinion veterinary practices across the United Kingdom
- Evaluation and test sets designed with distinct clinic distributions to measure model robustness