AAbAAC is a manually annotated corpus for information extraction in the biomedical domain of autoimmunity. The repository hosts:
- the annotated corpus
- the annotation guidelines
- the code used for named entity recognition (NER) experiments with this corpus
The corpus was constructed from 115 PubMed titles/abstracts selected for relevance to autoimmunity. It includes annotations for 5 entity types and 10 relation types, and each text was annotated by 2 annotators before adjudication into a final version. In the accompanying paper, the corpus is used to evaluate dictionary-based and neural NER approaches, including GLiNER fine-tuning.