Table 3 

Results of UDapter and UDify (Kondratyuk and Straka 2019) trained separately for POS tagging, morphological tagging, and NER on 13 high-resource languages. hr-avg is calculated over all 13 training languages, where ja and kr are removed from morphological tagging.

areneufihehiitjakorusvtrzhhr-avg
POS Tagging (Accuracy) 
multi-udify 96.3 96.6 94.5 96.4 96.5 96.7 98.3 96.8 95.3 98.5 97.6 93.4 93.8 96.1 
udapter 96.8 97.0 95.7 97.3 97.1 97.4 98.3 97.0 96.5 98.9 98.4 95.1 95.1 97.0 
 
Morphological Tagging (F1) 
multi-udify 95.7 97.5 92.0 94.7 94.9 97.0 98.9 – – 97.6 97.6 90.9 98.9 95.9 
udapter 96.8 98.0 95.4 96.7 96.2 97.7 99.2 – – 98.3 98.3 94.8 99.2 97.3 
 
NER (F1) 
multi-udify 87.9 84.1 91.3 90.9 84.7 86.8 91.3 68.8 87.3 88.6 94.4 91.9 78.8 86.7 
udapter 89.4 85.4 92.7 92.4 86.7 89.5 92.4 71.9 88.8 89.8 95.8 93.1 81.3 88.4 
areneufihehiitjakorusvtrzhhr-avg
POS Tagging (Accuracy) 
multi-udify 96.3 96.6 94.5 96.4 96.5 96.7 98.3 96.8 95.3 98.5 97.6 93.4 93.8 96.1 
udapter 96.8 97.0 95.7 97.3 97.1 97.4 98.3 97.0 96.5 98.9 98.4 95.1 95.1 97.0 
 
Morphological Tagging (F1) 
multi-udify 95.7 97.5 92.0 94.7 94.9 97.0 98.9 – – 97.6 97.6 90.9 98.9 95.9 
udapter 96.8 98.0 95.4 96.7 96.2 97.7 99.2 – – 98.3 98.3 94.8 99.2 97.3 
 
NER (F1) 
multi-udify 87.9 84.1 91.3 90.9 84.7 86.8 91.3 68.8 87.3 88.6 94.4 91.9 78.8 86.7 
udapter 89.4 85.4 92.7 92.4 86.7 89.5 92.4 71.9 88.8 89.8 95.8 93.1 81.3 88.4 
View Large
Close Modal
This Feature Is Available To Subscribers Only

or Create an Account

Close Modal
Close Modal
This site uses cookies. By continuing to use our website, you are agreeing to our privacy policy. No content on this site may be used to train artificial intelligence systems without permission in writing from the MIT Press.
Accept