You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
From the UDPipe point of view, XPOS is just another tag to predict and the predicted values are based on whatever was used for XPOS in the training data.
In CoNLL-U, XPOS is a treebank-specific PoS tag, so there is no unified reference. It is not needed in UD (there are treebanks without XPOS). Usually, it contains legacy pre-UD pos tags. For example, English uses PennTB PoS tags, Czech uses PDT tags - in both cases these XPOS tags encode not only the part of speech, but also other morphological features, which are encoded also in the CoNLL-U FEATS column. For other languages/treebanks, search their documentation (at least, the metadata should contain info such as manual native).
Is there a reference we can find what the abbreviations returned under the column
xpos
mean for English?I can find the ones for
upos
at this link (https://universaldependencies.org/u/pos/index.html); however, the ones underxpos
require a lot of guesswork.The text was updated successfully, but these errors were encountered: