One-shot learning of generative speech concepts
Under review
Genearting new examples
The links below provides examples (from humans and machines) of one-shot speech generation for new Japanese words. You can listen to the synthesized examples produced by:
-
Humans
(naive English speakers)
-
HSMM 1
(model with phone-like units learned from a Japanese corpus)
-
HSMM 2
(model with phone-like units learned from an English corpus).
Details
- These tasks are the various conditions of the Experiment 2 "replication".
- The first trial is a practice trial with English words.
- Actual participants performed 21 trials and had to pass an instructions quiz.