JHU Kaldi system for Arabic MGB-3 ASR challenge using diarization, audio-transcript alignment and transfer learning.
Vimal ManoharDaniel PoveySanjeev KhudanpurPublished in: ASRU (2017)
Keyphrases
- transfer learning
- manifold alignment
- learning tasks
- arabic handwriting recognition
- knowledge transfer
- labeled data
- reinforcement learning
- speaker identification
- automatic speech recognition
- cross domain
- multimedia
- transfer knowledge
- broadcast news
- semi supervised learning
- machine learning
- word level
- text categorization
- active learning
- collaborative filtering
- speech recognition
- text classification
- learning algorithm
- machine learning algorithms
- domain adaptation
- noisy environments
- speech corpus
- feature extraction
- cross lingual
- speech signal
- target domain
- transferring knowledge
- data mining
- word recognition
- multi task
- information retrieval
- domain specific
- text mining
- knowledge representation
- high dimensional