Figure 4
From: Pooling annotated corpora for clinical concept extraction
The plots A, B and C show the F1-score, precision and recall of the taggers respectively. The horizontal axis indicates the size of the MCR subsets used in the cross-validation. In each plot the colors of the horizontal lines correspond to the corpus used in the cross-validation experiment, viz. MCR, MCR + i2b2 (MCR with training fraction supplemented by i2b2) and MCR + i2b2C (MCR with training fraction supplemented by curated i2b2). The two types of evaluation methods are represented using different line styles: Exact (E) with solid lines and Overlap (O) with dashed lines.