Construction of Chinese Electronic Medical Record Entity and Entity Relation Annotation Corpus for Pediatric Bronchopneumonia Disease

This study investigated the construction of a corpus for annotating named entities and entity relationships in Chinese electronic medical records (EMRs) of pediatric bronchopneumonia, addressing the current lack of annotated data in this area. Guided by experts in the medical field, we developed an annotation system and detailed guidelines specifically tailored to named entities and entity relationships related to pediatric bronchopneumonia. Under the guidance of these system and guidelines, three rounds of manual annotation and correction were performed, resulting in the annotation of over 740,000 characters of pediatric bronchopneumonia EMR texts, including named entities and entity relationships. As a result, a corpus for pediatric bronchopneumonia EMRs, containing a total of 48,756 named entities and 9,558 entity relationships, was constructed. The inter-annotator agreement for the annotated data reached 98%. This work provides reliable data support for the study of named entity recognition and information extraction in Chinese EMRs, and is of significant importance for medical knowledge mining.