The performance of automatic speech recognition (ASR) systems has advanc...
Recent research using pre-trained transformer models suggests that just ...
Deep acoustic models represent linguistic information based on massive
a...
Many archival recordings of speech from endangered languages remain
unan...
For many (minority) languages, the resources needed to train large model...
For languages with insufficient resources to train speech recognition
sy...
Variation in speech is often represented and investigated using phonetic...