Skip to Main Content
Table 9

Adaptation using the preposition error correction task as an example. Models are trained on different sizes of native (WikiNYT) and learner data suing the AP algorithm. Each cell shows the F1 score. The same results are presented in Figure 3.

Amount of native training dataAmount of learner training data
03K32K
14.62 26.73 
1M 22.01 23.92 23.30 
5M 24.32 26.01 26.62 
13M 26.43 29.48 29.73 
Amount of native training dataAmount of learner training data
03K32K
14.62 26.73 
1M 22.01 23.92 23.30 
5M 24.32 26.01 26.62 
13M 26.43 29.48 29.73 
Close Modal

or Create an Account

Close Modal
Close Modal