Knowledge Distillation for Neural Transducer-based Target-Speaker ASR: Exploiting Parallel Mixture/Single-Talker Speech Data.
Takafumi MoriyaHiroshi SatoTsubasa OchiaiMarc DelcroixTakanori AshiharaKohei MatsuuraTomohiro TanakaRyo MasumuraAtsunori OgawaTaichi AsamiPublished in: INTERSPEECH (2023)
Keyphrases
- data sets
- data analysis
- raw data
- data mining techniques
- knowledge base
- training data
- speech recognition
- automatic speech recognition
- database
- multiple sources
- data collection
- knowledge discovery
- knowledge acquisition
- domain knowledge
- prior knowledge
- domain experts
- human experts
- data points
- knowledge management
- data sources
- data processing
- expert knowledge
- audio visual
- neural network
- input data
- high dimensional
- expert systems