Improvements to Embedding-Matching Acoustic-to-Word ASR Using Multiple-Hypothesis Pronunciation-Based Embeddings.
Hao YenWoojay JeonPublished in: CoRR (2022)
Keyphrases
- speech recognition
- multiple hypothesis
- speech recognition systems
- automatic speech recognition
- spontaneous speech
- speech recognizers
- vector space
- prosodic features
- speech recognizer
- acoustic models
- speaker independent
- word recognition
- speech sounds
- acoustic features
- spoken language
- word error rate
- speech signal
- pattern matching
- conversational speech
- string matching
- euclidean space
- matching algorithm
- co occurrence
- language model
- n gram
- hidden markov models
- particle filter
- recognition errors
- graph matching
- speech synthesis
- spoken document retrieval
- image matching
- human machine interaction
- feature points
- low dimensional
- dimensionality reduction
- handwriting recognition
- word segmentation
- latent space
- low dimensional spaces