Skip to content

Latest commit

 

History

History
60 lines (55 loc) · 5.31 KB

evaluation.md

File metadata and controls

60 lines (55 loc) · 5.31 KB

Evaluation results

Stories dataset

Precision Recall F-score Map
Most frequent class 0.31 0.5 0.383 0.623
SVM TFIDF: targetExp 0.911 0.893 0.902 0.928
SVM WordEmb: targetExp 0.927 0.919 0.923 0.954
BERTClassifier: targetExp 0.951 0.948 0.949 0.985
SVM TFIDF: context3w 0.734 0.739 0.737 0.859
SVM WordEmb: context3w 0.758 0.742 0.75 0.876
BERTClassifier: context3w 0.931 0.926 0.929 0.978
SVM TFIDF: context3wmasked 0.674 0.677 0.675 0.804
SVM WordEmb: context3wmasked 0.674 0.678 0.676 0.809
BERTClassifier: context3wmasked 0.855 0.852 0.854 0.951
SeqModel: LSTM 0.952 0.948 0.95 0.949
MaskPredict: BERT-sent, contemporary, weighted, wsdbert 0.739 0.703 0.72 0.848
MaskPredict: BERT-both, contemporary, weighted, wsdbert 0.839 0.774 0.806 0.892
MaskPredict: BERT-both, before1850, weighted, wsdbert - - - -
MaskPredict: BERT-both, timeSensitive, weighted, wsdbert - - - -

19thC Machines dataset (animacy)

Precision Recall F-score Map
Most frequent class 0.336 0.5 0.402 0.318
SVM TFIDF: targetExp 0.696 0.713 0.704 0.474
SVM WordEmb: targetExp 0.694 0.711 0.702 0.499
BERTClassifier: targetExp 0.698 0.715 0.706 0.51
SVM TFIDF: context3w 0.688 0.71 0.699 0.651
SVM WordEmb: context3w 0.728 0.531 0.614 0.481
BERTClassifier: context3w 0.695 0.721 0.708 0.721
SVM TFIDF: context3wmasked 0.592 0.6 0.596 0.498
SVM WordEmb: context3wmasked 0.518 0.52 0.519 0.339
BERTClassifier: context3wmasked 0.687 0.696 0.692 0.603
SeqModel: LSTM 0.697 0.719 0.708 0.482
MaskPredict: BERT-sent, contemporary, weighted, wsdbert 0.719 0.742 0.73 0.74
MaskPredict: BERT-both, contemporary, weighted, wsdbert 0.758 0.778 0.768 0.795
MaskPredict: BERT-both, timeSensitive, weighted, wsdbert 0.758 0.775 0.766 0.777
MaskPredict: BERT-both, before1850, weighted, wsdbert 0.799 0.773 0.786 0.784

19thC Machines dataset (humanness)

Precision Recall F-score Map
Most frequent class 0.434 0.5 0.465 0.155
SVM TFIDF: targetExp 0.434 0.499 0.464 0.122
SVM WordEmb: targetExp 0.525 0.548 0.536 0.123
BERTClassifier: targetExp 0.55 0.588 0.568 0.201
SVM TFIDF: context3w 0.545 0.563 0.554 0.213
SVM WordEmb: context3w 0.484 0.486 0.485 0.182
BERTClassifier: context3w 0.538 0.556 0.547 0.177
SVM TFIDF: context3wmasked 0.521 0.54 0.53 0.219
SVM WordEmb: context3wmasked 0.539 0.586 0.562 0.16
BERTClassifier: context3wmasked 0.6 0.671 0.633 0.267
SeqModel: LSTM 0.563 0.544 0.554 0.237
MaskPredict: BERT-sent, contemporary, weighted, wsdbert 0.553 0.612 0.581 0.204
MaskPredict: BERT-both, contemporary, weighted, wsdbert 0.553 0.613 0.582 0.236
MaskPredict: BERT-both, timeSensitive, weighted, wsdbert 0.587 0.657 0.62 0.249
MaskPredict: BERT-both, before1850, weighted, wsdbert 0.593 0.664 0.627 0.277