Improving speaker turn embedding by crossmodal transfer learning from face embedding.
Nam LeJean-Marc OdobezPublished in: CoRR (2017)
Keyphrases
- transfer learning
- learning tasks
- vector space
- latent space
- labeled data
- knowledge transfer
- manifold alignment
- transfer knowledge
- face images
- domain adaptation
- text classification
- cross domain
- reinforcement learning
- machine learning algorithms
- semi supervised learning
- text categorization
- unlabeled data
- domain specific
- semi supervised
- natural language
- face recognition