Monolingual Word Alignment on MultiMWA-Newsela
95.2Precision (P)PipelineAligner
Evaluation Results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| PipelineAlignerTraining Dataset=MultiMWA-MTRef Sure+Poss2021.06 | 95.2 | 69.4 | 80.3 | 10 | 96.9 | 64.4 | 95.3 | 25.4 | 96.1 | 36.5 | |
| Neural semi-CRF AlignerTraining Dataset=MultiMWA-MTRef Sure+Poss2021.06 | 89.4 | 85 | 87.2 | 21.6 | 96.7 | 76.1 | 98.4 | 66.5 | 97.6 | 71 | |
| Neural CRF AlignerTraining Dataset=MultiMWA-MTRef Sure+Poss2021.06 | 88.2 | 85 | 86.6 | 15.6 | 95.3 | 72.3 | 99 | 66.3 | 97.1 | 69.1 | |
| JacanaTokenTraining Dataset=MultiMWA-MTRef Sure+Poss2021.06 | 85.5 | 74.9 | 79.8 | 11 | 91.2 | 60.1 | 97.5 | 39.7 | 94.3 | 47.9 | |
| QA-based AlignerTraining Dataset=MultiMWA-MTRef Sure+Poss2021.06 | 84.8 | 87.9 | 86.2 | 16.2 | 95.3 | 69.4 | 99.1 | 71.4 | 97.1 | 70.4 | |
| JacanaPhraseTraining Dataset=MultiMWA-MTRef Sure+Poss2021.06 | 84.3 | 75 | 79.4 | 8.2 | 91.3 | 53.9 | 97.4 | 38.6 | 94.3 | 45 |