Final Report on the German Clinical Reference Corpus 3000PA

Area of Science:

• Medical Informatics
• Computational Linguistics
• Natural Language Processing

Background:

• The Medical Informatics Initiative (MII) is a large-scale German research program.
• Development of robust data and software infrastructure is crucial for German-language clinical natural language processing.

Purpose of the Study:

• To report on the development of a national clinical reference corpus (3000PA) and complementary sharable corpora.
• To support the advancement of German clinical natural language processing.

Main Methods:

• Developed 3000PA, a national clinical reference corpus from three university hospital patient records.
• Annotated 3000PA with semantic layers: medical named entities, relations, certainty, and negation.
• Created three sharable corpora: JSYNCC, GGPONC, and GRASCCO.

Main Results:

• The 3000PA corpus contains detailed semantic annotations.
• Three sharable corpora (JSYNCC, GGPONC, GRASCCO) complement 3000PA.
• Combined corpora (3000PA, JSYNCC, GRASCCO) feature approximately 2.1 million metadata points.

Conclusions:

• The developed corpora provide a valuable resource for German clinical natural language processing research.
• These resources enhance the data and software infrastructure for medical informatics in Germany.