Ig Nobel classification on Custom labeled Ig Nobel (k=5 cross-validation)
92.2AccuracySciBERT^f
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| SciBERT^ftraining=fine-tuned, features=combined with literature-based features2021.06 | 92.2 | 91.9 | 92.6 | |
| SciBERTtraining=fine-tuned2021.06 | 91 | 91.1 | 91.1 | |
| BERTtraining=fine-tuned2021.06 | 90.4 | 90.6 | 89.3 | |
| BERT^ftraining=fine-tuned, features=combined with literature-based features2021.06 | 90 | 89.9 | 90.2 | |
| Iggydescription=literature-based features model2021.06 | 89.7 | 90.1 | 89.3 | |
| LRmethod=logistic regression, input=bag of words (BoW)2021.06 | 78.1 | 75.4 | 83.7 | |
| RFmethod=random forest, input=bag of words (BoW)2021.06 | 76.1 | 74.6 | 79.6 |