Humanized Language Generation and Identification
Humanized Language Generation and Identification
Professor Tao Jianhua
human-friendly language generation
1. Background introduction
-
Machine Learning Advances Rapidly
- Promote the development and application of language and image fields
-
limitation
- A large amount of structured data is required, and the performance of small data is poor
-
voice instance
- A practical language recognition system requires a large amount of labeled data
- It takes more than ten hours for a single person to synthesize a language close to a real person, but there is very little personalized language data
-
challenge
- In many cases, it is difficult to obtain a large amount of speech annotation data
- Minor languages
- small data modeling
- Leverage 'Big Data' -> 'Small Data'
- Leverage 'Big Data' -> 'Small Data'
2. Transfer Learning
3. Speech generation related work
Speech Generation Reporting Work
human language identification
1. Language confrontation - generating speech discrimination
2. Research status at home and abroad
3. Propose work