Sentiment Analysis on NSMC (dev)
89.59AccuracyBERT
Evaluation results
| Method | Links | |
|---|---|---|
| BERTTokenization=MorSubword, Vocab Size=32K2026.04 | 89.59 | |
| BERTTokenization=Character, Vocab Size=2K2026.04 | 89.03 | |
| BERTTokenization=Subword, Vocab Size=32K2026.04 | 89.03 | |
| KOMBO_JamoTokenization=Jamo, Vocab Size=1702026.04 | 88.71 | |
| BERTTokenization=Jamo, Vocab Size=1702026.04 | 87.84 | |
| KOMBO_StrokeTokenization=Stroke, Vocab Size=1302026.04 | 87.65 | |
| BERTTokenization=Morpheme, Vocab Size=32K2026.04 | 87.55 | |
| BERTTokenization=Stroke, Vocab Size=1302026.04 | 87.02 | |
| BERTTokenization=Cji, Vocab Size=1362026.04 | 86.85 | |
| KOMBO_BTSTokenization=BTS, Vocab Size=1122026.04 | 86.68 | |
| KOMBO_CjiTokenization=Cji, Vocab Size=1362026.04 | 86.55 | |
| BERTTokenization=BTS, Vocab Size=1122026.04 | 86.14 | |
| BERTTokenization=Word, Vocab Size=64K2026.04 | 74.91 |