ON-TRAC Consortium Systems for the IWSLT 2022 Dialect and Low-resource Speech Translation Tasks

20662

Knowledge Triples

Source Entity	Relationship	Target Entity
PhoneticAnalysis Method	improves	Speech Translation Task
Self-Supervised Models Method	improves	Speech Translation Task
Tamasheq French dataset Material	leverages	Wav2Vec 2.0 Material
Tamasheq French dataset Material	uses	Conformer speech translation architecture Method
Tunisian Arabic English dataset Material	comparesWith	Transfer Learning Method
Tunisian Arabic English dataset Material	leverages	Wav2Vec 2.0 Material
Wav2Vec 2.0 Material	usedFor	Speech Translation Task