Part-of-Speech Tagging on TWEEBANK V2 (test)
95.8AccuracyACE
Evaluation results
| Method | Links | |
|---|---|---|
| ACEFine-tuning=true, Trained on both train and development set=true2020.10 | 95.8 | |
| XLM-R-largeNormalization strategy=soft2020.05 | 95.5 | |
| XLM-RFine-tuning=true, Trained on both train and development set=true2020.10 | 95.4 | |
| HuggingFace-BERTweetTraining Data=TB2 + UD English EWT2022.01 | 95.38 | |
| HuggingFace-BERTweetTraining Data=TB22022.01 | 95.21 | |
| BERTweetNormalization strategy=soft2020.05 | 95.2 | |
| Nguyen et al.2020.10 | 95.2 | |
| BERTweet2022.01 | 95.2 | |
| XLM-R-largeNormalization strategy=hard2020.05 | 95.1 | |
| RoBERTa-largeNormalization strategy=soft2020.05 | 94.9 | |
| XLM-R-baseNormalization strategy=soft2020.05 | 94.7 | |
| BERTweetNormalization strategy=hard2020.05 | 94.7 | |
| Owoputi et al., 2013 (CRF)Tokenization=gold-standard, Inference strategy=CRF, Training Data=UD_English-EWT + TWEEBANK V22018.04 | 94.6 | |
| RoBERTa-largeNormalization strategy=hard2020.05 | 94.6 | |
| ARKtaggerNormalization strategy=hard, Additional training data=UD_English-EWT (+c)2020.05 | 94.6 | |
| Owoputi et al.2020.10 | 94.6 | |
| Owoputi et al. (2013)Decoding=CRF2022.01 | 94.6 | |
| XLM-R-baseNormalization strategy=hard2020.05 | 94.3 | |
| spaCy-XLM-RoBERTaTraining Data=TB22022.01 | 93.9 | |
| spaCy-XLM-RoBERTaTraining Data=TB2 + UD English EWT2022.01 | 93.75 | |
| Owoputi et al., 2013 (greedy)Tokenization=gold-standard, Inference strategy=greedy, Training Data=UD_English-EWT + TWEEBANK V22018.04 | 93.7 | |
| RoBERTa-baseNormalization strategy=soft2020.05 | 93.7 | |
| Owoputi et al. (2013)Decoding=greedy2022.01 | 93.7 | |
| StanzaTraining Data=TB2 + UD English EWT2022.01 | 93.53 | |
| RoBERTa-baseNormalization strategy=hard2020.05 | 93.5 | |
| StanzaTraining Data=TB22022.01 | 93.2 | |
| Gui et al.2020.10 | 92.8 | |
| Ma and Hovy, 2016Tokenization=gold-standard, Model=Neural network, Training Data=UD_English-EWT + TWEEBANK V22018.04 | 92.5 | |
| BiLSTM-CNN-CRFNormalization strategy=hard, Additional training data=UD_English-EWT (+c)2020.05 | 92.5 | |
| Ma and Hovy (2016)2022.01 | 92.5 | |
| Stanford CoreNLPTokenization=gold-standard, Training Data=UD_English-EWT + TWEEBANK V22018.04 | 90.6 | |
| Stanford CoreNLP2022.01 | 90.6 | |
| spaCyTraining Data=TB2 + UD English EWT2022.01 | 88.84 | |
| FLAIRTraining Data=TB2 + UD English EWT2022.01 | 88.19 | |
| FLAIRTraining Data=TB22022.01 | 87.85 | |
| spaCy-BERTweetTraining Data=TB22022.01 | 87.61 | |
| spaCyTraining Data=TB22022.01 | 86.72 | |
| spaCy-BERTweetTraining Data=TB2 + UD English EWT2022.01 | 86.31 |