Capsule Routing for Named Entity Recognition
-> Conll2003: dev, train, test
-> Rotten Tomatoes Dataset: train, test
| Model | CoNLL-2003 | Rotten Tomatoes Dataset | Validation Accuracy |
|---|---|---|---|
| CapsNet | ✔️ | ✖️ | 96.47 |
| CapsNetGRU+FastText | ✖️ | ✔️ | 85.86 |
| CapsNetGRU+Toxic | ✖️ | ✔️ | 85.74 |
| BLSTM + CBOW | ✔️ | ✖️ | 89.65 |
| BLSTM + Skipgram | ✔️ | ✖️ | 89.78 |
| BLSTM+FastText | ✔️ | ✖️ | 89.72 |
| BLSTM+Glove6B50d | ✔️ | ✖️ | 97.17 |
| BLSTM+Glove6B200d | ✔️ | ✖️ | 97.52 |
| BLSTM+Glove6B300d | ✔️ | ✖️ | 97.60 |
| BLSTM+Glove42B300d | ✔️ | ✖️ | 97.75 |
| BLSTM+Glove840B300d | ✔️ | ✖️ | 97.01 |