How to restore case during translation when using case_feature?


I used case_feature during preprocessing as suggested to optimize the vocabulary. The translated sentences have the cases specified alongside them for each word. Is it possible to directly restore the correct casing during translation ? I now have something like this

a | c dog | l is | l

I would like to get

A dog is

as the final output


You need to run the detokenization script with -case_feature flag.

1 Like