SLT Track

The goals of the SLT track is to transcribe and translate TED and TEDx lectures. The audio of the talks is segmented into sentence like chunks and the use of the segmentation is mandatory. Reference ASR outputs are provided to all participants. These outputs are either the best ASR system available for a set or a combination of systems from multiple partners.

For the evaluation the ASR output provided will be a ROVER combination of all outputs.

• Language directions: English->{Arabic, Chinese (Mandarin), German, Italian}, {German, Italian}->English

• Development Data:

  • English->French dev data: dev2010, dev2010.1best, tst2010, tst2010.1best

  • English->{German,Italian,Arabic,Chinese} dev data: dev2010, dev2010.1best

  • German->English dev data: dev2012, dev2012.1best

  • Italian->English dev data: dev2014, dev2014.1best

• Evaluation Data:

  • English->{French,German,Italian,Arabic,Chinese}: tst2014

  • German->English: tst2014

  • Italian->English: tst2014