🐛 Bug
Audio segments from speech data in example speech translation task (IWSLT17 dev) are not correctly associated with reference data.
Only the first TED talk audio segments are correctly aligned to the reference... playing the audio segments related to any other talks (from # 3 / 10 / 887 ( 153 / 887 ) onwards on page 16 of the task using the defaults) results in the segments of the first TED talk audio being played rather than the segments specified in the task directory speech_translation_iwslt17_dev/src_0.zip/source.txt
To reproduce
Get the example speech task data (IWSLT7 dev)
$ bash get_example_data.sh speech_translation_iwslt17_dev
Start the server and navigate to :
http://127.0.0.1:5000/view?t=speech_translation_iwslt17_dev&m=&q=&p_sz=10&p_no=16&s=0&s_metric=
Play the audio segments: first two on this page will be correctly associated with reference text, from # 3 / 10 / 887 ( 153 / 887 ) onwards they are not.
🐛 Bug
Audio segments from speech data in example speech translation task (IWSLT17 dev) are not correctly associated with reference data.
Only the first TED talk audio segments are correctly aligned to the reference... playing the audio segments related to any other talks (from # 3 / 10 / 887 ( 153 / 887 ) onwards on page 16 of the task using the defaults) results in the segments of the first TED talk audio being played rather than the segments specified in the task directory speech_translation_iwslt17_dev/src_0.zip/source.txt
To reproduce
Get the example speech task data (IWSLT7 dev)
Start the server and navigate to :
http://127.0.0.1:5000/view?t=speech_translation_iwslt17_dev&m=&q=&p_sz=10&p_no=16&s=0&s_metric=
Play the audio segments: first two on this page will be correctly associated with reference text, from # 3 / 10 / 887 ( 153 / 887 ) onwards they are not.