Compositional Sequence Labeling Models for Error Detection in Learner Writing

Benchmark Model Rank Results
grammatical-error-detection-on-conll-2014-a1Bi-LSTM (trained on FCE)–F0.5: 16.4
grammatical-error-detection-on-conll-2014-a1Bi-LSTM (unrestricted data)–F0.5: 34.3
grammatical-error-detection-on-conll-2014-a2Bi-LSTM (trained on FCE)–F0.5: 23.9
grammatical-error-detection-on-conll-2014-a2Bi-LSTM (unrestricted data)–F0.5: 44.0
grammatical-error-detection-on-fceBi-LSTM–F0.5: 41.1