Equivalence Matching Results for SNOMED-FMA (Body)
 
Unsupervised (No Train Maps)
Semi-supervised (Optional 30% Train Maps)
⇕System ⇕Precision ⇕Recall ⇕F-score ⇕MRR ⇕Hits@1 ⇕Use Train Maps ⇕Precision ⇕Recall ⇕F-score ⇕MRR ⇕Hits@1
AMD∗ 0.890 0.633 0.740 ✔ 0.861 0.709 0.778
BERTMap‡ 0.979 0.662 0.790 0.944 0.920 ✔ 0.970 0.669 0.792 0.965 0.947
BERTMapLt‡ 0.979 0.655 0.785 0.892 0.865 ✘ 0.970 0.662 0.787 0.892 0.865
LogMap† 0.744 0.407 0.526 ✘ 0.673 0.411 0.511
LogMapBio† 0.827 0.577 0.680 ✘ 0.770 0.577 0.660
LogMapLt† 0.970 0.542 0.696 ✘ 0.958 0.542 0.693
Matcha† 0.887 0.502 0.641 ✘ 0.846 0.502 0.630
Matcha-DL∗ 0.960 0.602 0.740 0.918 0.908 ✔ 0.959 0.825 0.887 0.918 0.908
OLaLa∗ 0.270 0.348 0.304 ✘ 0.202 0.339 0.253
SORBETMtch†∗ 0.618 0.749 0.677 ✔ 0.794 0.704 0.746

Note: The BERTMap model for this task does not apply mapping repair on this task as the reasoning was not able to be completed.

Note: Mixed results for SORBETMtch, where unsupervised results come from MELT and semi-supervised results come from direct result submission.