Closed kharaldsson closed 1 year ago
Let's discuss it tomorrow. I wonder whether we just use textblob. That could be the extent of our adaptation. Translate everything to english. Train on the combined en and es training set (post-translation). Perhaps even pre-translate test set.
Per discussion today, @alex-321 will develop a script that takes a Spanish file and outputs a deep_translator translated version of it. Such files will be written to data/train/es2en
, data/dev/es2en
, data/test/es2en
.
FYSA running via CLI:
Packages and models tried:
ProtocolError: Invalid input ConnectionInputs.RECV_PING in state ConnectionState.CLOSED
when translating dev set