Five learners were involved in this experiment. They are all male to
keep balance with the speakers of the speech database. A native
speaker's scores are used to decide an adequate threshold value. Each
learners produced just one repetition for the data set of 13
sentences, which are taken from a text book of Japanese
language. Experiments were carried out for the continous speech utterances
containing coarticulation. The T-set consists of 13 sentences as follows:
- moshimoshi, korekaradokoeikimasuka
(Hello, where are you going now?)
- watashi, kyootoeikimasu
(Me? To Kyoto.)
- nanideikimasuka
(How are you going there?)
- shiNkaNseNdeikimasu
(I go there by shinkansen.)
- konogaikokujiNtoorokushiNseishookaitekudasai
(Would you please fill out this foreigner registration form?) - zeNbukakuNdesuka
(Should I fill in all?)
- hai, konohutowakunonakaokaitekudasai
(Yes, please fill out the items in a bold line.)
- chotokakikatagawakaranaiNdesuga
(Sorry, I don't know how to fill it out.)
- kokoninamaeokaitekudasai
(Write your name here, please.)
- romajideiiNdesuka
(Is is okay to write it in a roman character?)
- romajidekekoodesu
(You may write it in a roman character.)
- kokoninaniokakuNdesuka
(What should I write it down here?)
- kokoniwajuushookaitekudasai
(Would you please write down your address here.)
Next: Automatic Scoring
Up: 3.6.3 Task 3 :
Previous: 3.6.3 Task 3 :
Jo Chul-Ho
Wed Oct 13 17:59:27 JST 1999