Extract features for many target translations per source sentence

lspecia / quest

Pascal2 Harvest project QuEst

14 stars 15 forks source link

Extract features for many target translations per source sentence #4

Open lefterav opened 11 years ago

lefterav commented 11 years ago

The current setup only allows the extraction of features for one source and one target. This is quite hard-coded on the respective class. For advanced experiments (e.g. sentence ranking QTLPWP2) we would be interested in use cases where there is only one source sentence and many respective targets

lefterav commented 11 years ago

I am working on this on branch sentence_level. This will work by reading and writing XML files, instead of tabbed plain text files, in order to allow a flexible number of target sentences per sentence