As new pretranslations are made in machine.py in #662, they will need to be fed up through machine and serval and saved in the database. The option should be added in build jobs to always generate the alignments. The pretranslation structures should assume that new data will always have the alignments (and tokenized versions and source), but that existing data may not have those fields.
As new pretranslations are made in machine.py in #662, they will need to be fed up through machine and serval and saved in the database. The option should be added in build jobs to always generate the alignments. The pretranslation structures should assume that new data will always have the alignments (and tokenized versions and source), but that existing data may not have those fields.