Herb Disease relation corpus

The corpus is unique and will be an important resource for oriental medicine text mining system.

Find Out More

About corpus

In this research, we define herb and disease relation as treatment of disease and cause of disease. We annotate 1,013 relations from 175 abstract. In order to verify the effectiveness of the corpus by applying to Turku event extraction system. F-score with 5-fold cross validation was 80.26.

Download corpus!


1,013 relations

From 176 PubMed abstracts.

2,032 entities

1,006 herb entities and 1,016 disease entities.

81 to 99 intert-annoatator agreements

by two curators.

F-score = 80.28

relation prediction using Turku Event Extraction System.