Pooling annotated corpora for clinical concept extraction

Table 2 Overlap of annotations in the corpora

Annotation set	Exact match	Ignoring one start word	Ignoring one word
MCR	42.9/43.7	55.1/55.7	56.0/56.7
i2b2	22.6/25.2	32.6/33.6	33.6/34.7

Each cell in the table gives the percentage of annotations that matched with the other corpus before and after automated curation of the i2b2 corpus to partially rectify the annotation differences.

ISSN: 2041-1480