CMU Multilingual NLP 2020 - Low Resource ASR

Learn about applying Automatic Speech Recognition (ASR) systems to low- or no-resource languages in this 44-minute lecture from CMU's Multilingual Natural Language Processing course. Explore topics such as ASR without lexicons or transcribed data, cross-lingual phone recognition, text-to-speech without text, iterative decoding in German and English, speech translation, intent discovery from acoustics, and keyword matching. Gain insights into the challenges and innovative approaches for developing ASR systems in resource-constrained environments, presented by Alan Black as part of the CS11-737 course at Carnegie Mellon University.

Intro
ASR with no Lexicon
ASR without transcribed data
CMU Wildemess
Cross-Lingual Phone Recognition
TTS without the Text
Iterative Decoding: German
Iterative Decoding: English
Speech Translation
Intent Discovery from Acoustics
Keyword Matching
ASR no data: Discussion Point