This resource is the result of ACLR's research project Multi-modal Chinese Interlanguage Speech Corpus for Intelligent Pronunciation Teaching. The project manager is Zhang Jinsong, chief expert of ACLR and Professor of BLCU.
The project mainly includes the following resource achievements:
1. Chinese Interlanguage Pronunciation Database
The total number of speakers in this database is 894, from 43 countries and regions, 16 countries with more than 10 speakers and 11 countries with more than 20 speakers. The length of the file exceeds 400 hours. The phase I and II data collected from 675 learners with different native languages, have been completed with phonetic annotation. The database also contains recordings of 82 native Chinese speakers which will be used as standard pronunciation for comparative research.
2. Auditory Perception Database for Acquisition of Chinese as a Second Language
This database is aimed at solving difficulties in learning Chinese tones, initials (l/ r, z, c, s/zh, ch, sh, b, d, g/p, t, k), front and back nasal vowels, focus stress, etc. Participants came from 15 countries including China, Japan, South Korea, Pakistan, Vietnam, Laos, Bangladesh, Thailand, Russia, and Italy. The total number of foreign participants exceeded 300, and the number of Chinese participants was 390.
3. Database for Speech Pronunciation Physiological Data and Videos
It contains the EMA data of speech and pronunciation organs of 4 native Chinese speakers, the oral muscle training videos, and the oral pronunciation videos of 2D and 3D animations. The database is mainly used for video feedback for pronunciation correction software.
4. Multilingual Parallel Speech Database
Based on 301 common Chinese expressions, a small number of speakers’ parallel speeches in 17 languages are collected, including English, Japanese, French, German, Russian, Mongolian, Spanish, Portuguese, Korean, Arabic, and Urdu, Hindi, Malay, Thai, Vietnamese, Bengali, Filipino, etc.It can be used for L1 and L2 speech comparison research or for acoustic model transfer learning.
5. Pronunciation Development Corpus for Second Language Learners
The Corpus contains 21 hours of 28,000 pieces of voice data from 36 international students speaking 19 native languages, 780 pieces per capita. The data were collected when these students participate in a 6-week pronunciation learning task based on the SAIT Chinese APP. The data reveal learners’ pronunciation development during their learning process, and can be used to research on Chinese pronunciation acquisition from a time-varying perspective.
For more details of project achievements, please click:
http://yuyanziyuan.blcu.edu.cn/info/1066/2525.htm
