NoRefER: a Referenceless Quality Metric for Automatic Speech Recognition via Semi-Supervised Language Model Fine-Tuning with Contrastive Learning.
Kamer Ali YukselThiago Castro FerreiraGolara JavadiMohamed Al-BadrashinyAhmet GunduzPublished in: CoRR (2023)
Keyphrases
- language model
- speech recognition
- fine tuning
- automatic speech recognition
- word error rate
- semi supervised
- language modeling
- information retrieval
- supervised learning
- n gram
- document retrieval
- active learning
- information extraction
- quality metrics
- probabilistic model
- human perception
- retrieval model
- mixture model
- test collection
- query expansion
- wavelet transform
- hidden markov models
- natural language