Stance Detection on FNC-1 (test)
0.8123Weighted AccuracysMemNN
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| sMemNNTotal Parameters=110M, Trainable Parameters=105M, enrichment=BOW with TF.IDF2018.04 | 0.8123 | 0.5688 | 0.8857 | |
| sMemNNTotal Parameters=5.5M, Trainable Parameters=377.5K2018.04 | 0.7897 | 0.5675 | 0.8727 | |
| Gradient Boosting2018.04 | 0.752 | 0.4613 | 0.8632 | |
| sMemNNTotal Parameters=5.4M, Trainable Parameters=275.2K, Similarity type=dot product2018.04 | 0.7513 | 0.5021 | 0.8385 | |
| LSTM+CNNTotal Parameters=2.8M, Trainable Parameters=281.5K2018.04 | 0.6021 | 0.4033 | 0.6536 | |
| LSTMTotal Parameters=2.8M, Trainable Parameters=261.3K2018.04 | 0.5723 | 0.3723 | 0.6021 | |
| All-discuss2018.04 | 0.4389 | 0.0747 | 0.1757 | |
| CNN+LSTMTotal Parameters=4.2M, Trainable Parameters=361.5K2018.04 | 0.4202 | 0.2736 | 0.4854 | |
| CNNTotal Parameters=2.7M, Trainable Parameters=188.7K2018.04 | 0.4066 | 0.2444 | 0.4153 | |
| All-unrelated2018.04 | 0.3937 | 0.2096 | 0.722 |