Improvements to Embedding-Matching Acoustic-to-Word ASR Using Multiple-Hypothesis Pronunciation-Based Embeddings.
Hao YenWoojay JeonPublished in: ICASSP (2023)
Keyphrases
- speech recognition
- multiple hypothesis
- speech recognition systems
- spontaneous speech
- automatic speech recognition
- speech recognizers
- prosodic features
- vector space
- speaker independent
- speech recognizer
- particle filter
- word recognition
- speech sounds
- human machine interaction
- word error rate
- acoustic features
- conversational speech
- acoustic models
- hidden markov models
- noisy environments
- speech synthesis
- spoken document retrieval
- pattern matching
- string matching
- speech signal
- graph matching
- recognition errors
- image matching
- spoken language
- speaker identification
- manifold embedding
- binary codes
- low dimensional spaces
- distance measure
- n gram
- matching algorithm
- spelling correction
- broadcast news
- language learning
- word sense disambiguation
- similarity search
- language model
- latent space
- co occurrence
- speech retrieval
- semi supervised
- feature selection