Improving Identification of System-Directed Speech Utterances by Deep Learning of ASR-Based Word Embeddings and Confidence Metrics.
Vilayphone VilaysoukAmr Nour-EldinDermot ConnollyPublished in: ICASSP (2021)
Keyphrases
- deep learning
- automatic speech recognition
- speech sounds
- word error rate
- speech segments
- speech recognition
- spontaneous speech
- speech signal
- conversational speech
- unsupervised feature learning
- unsupervised learning
- machine learning
- speech retrieval
- broadcast news
- mental models
- noisy environments
- spoken language
- spoken document retrieval
- vocal tract
- deep architectures
- language model
- hidden markov models
- dimensionality reduction
- n gram
- weakly supervised
- data mining
- higher order
- bayesian networks
- computer vision