SubRegWeigh: Effective and Efficient Annotation Weighing with Subword Regularization

Benchmark Model Rank Results
named-entity-recognition-ner-on-conll-2003LUKE + SubRegWeigh (K-means)#4F1: 94.2
named-entity-recognition-ner-on-conll-2003RoBERTa + SubRegWeigh (K-means)#10F1: 93.81
named-entity-recognition-on-conllLUKE + SubRegWeigh (K-means)#1F1: 96.12
named-entity-recognition-on-conllRoBERTa + SubRegWeigh (K-means)#4F1: 95.45
named-entity-recognition-on-wnut-2017RoBERTa + SubRegWeigh (K-means)#2F1: 60.29
semantic-textual-similarity-on-mrpcRoBERTa + SubRegWeigh (K-means)#25Accuracy: 86.82%
sentiment-analysis-on-sst-2-binaryRoBERTa + SubRegWeigh (K-means)#27Accuracy: 94.84