Fig. 3
From: Similarity corpus on microbial transcriptional regulation
The progress of IRR through the consensus sessions. The chart shows the IRR measured using five different metrics. The IRR score is represented on the y-axis, and the results for the four sessions are chronologically displayed on the x-axis. IRR scores, in all metrics, improved in each subsequent consensus session. For example, the IRR measured using Gwet’s AC2 coefficients improved from 0.545 in the first session to 0.910 in the last one, that is, the annotators’ evaluations were much more homogeneous at the end of the consensus sessions