From where can one get ted_talks_langs.txt and to what do we have to pass it as an input?

[ekdnam]

In scripts/get-ted-talks-data.sh, the following line has been commented.

improving-zeroshot-nmt/scripts/get-ted-talks-data.sh

Line 14 in b1fcda2

# we use ./scripts/ted_talks_langs.txt to extract all langs-en pairs to ./data/ted_data

From where can we get that file, what is its utility, and to which module do we have to pass it as an input?

[surafelml]

My apologies for the delay! Please see the latest version

improving-zeroshot-nmt/scripts/get-ted-talks-data.sh

Lines 14 to 16 in 8385d29

	# use the pre-specified src-trg lang-pairs from ./scripts/ted_talks_langs.txt to extract parallel data from ./ted_data
	# for pairs other than it/ro-en, update ./scripts/ted_talks_langs.txt
	python $EXPDIR/scripts/ted_reader.py --lang-file $EXPDIR/scripts/ted_talks_langs.txt --data-path ./ted-data/