Skip to Main Content
Table 13

Adapting NB with data from the target language background and other language backgrounds. Numbers in the parentheses show the relative improvement of using adapted models vs. the native-trained models. All improvements of adapted models vs. native-trained models are statistically significant (McNemar's test, p < 0.001).

ErrorNative-trainedAdapted (other)Adapted (target)
PRF1PRF1 (rel. improv.)PRF1 (rel. improv.)
Article 28.06 41.49 33.47 31.17 40.65 35.28 (5.4%) 32.21 41.33 36.20 (8.2%) 
Prep. 18.41 24.97 21.19 23.25 23.82 23.63 (11.5%) 23.80 24.95 24.25 (14.4%) 
Verb agr. 28.52 46.51 35.32 32.73 43.78 37.44 (6.0%) 33.58 45.96 38.80 (9.9%) 
ErrorNative-trainedAdapted (other)Adapted (target)
PRF1PRF1 (rel. improv.)PRF1 (rel. improv.)
Article 28.06 41.49 33.47 31.17 40.65 35.28 (5.4%) 32.21 41.33 36.20 (8.2%) 
Prep. 18.41 24.97 21.19 23.25 23.82 23.63 (11.5%) 23.80 24.95 24.25 (14.4%) 
Verb agr. 28.52 46.51 35.32 32.73 43.78 37.44 (6.0%) 33.58 45.96 38.80 (9.9%) 
Close Modal

or Create an Account

Close Modal
Close Modal