| 研究課題/領域番号 |
17K02937
|
|---|
| 研究機関 | 高知工科大学 |
|---|
研究代表者 |
ダニエルズ ポール 高知工科大学, 共通教育教室, 教授 (50307245)
|
|---|
| 研究分担者 |
熊井 信弘 学習院大学, 付置研究所, 教授 (00248999)
|
|---|
| 研究期間 (年度) |
2017-04-01 – 2020-03-31
|
| キーワード | automated / scoring / speaking / voice recognition |
|---|
| 研究実績の概要 |
1. An investigative study was conducted to evaluate how accurate cloud-based voice recognition engines are at transcribing speech with Japanese learners of English. Results of the study were published in an international journal. The research was also presented at several domestic and international language conferences.
2. Version 1 of the speech assessment system was successfully developed and deployed at three universities. It is currently being used for individualized online speaking practice.
3. A set of online speaking tasks were created using content from the course textbooks in order to provide learners with individualized online speaking practice. This online practice allows learners to practice for their authentic speaking tasks such as presentations and interviews.
4. Our automated scoring algorithm was improved to more accurately evaluate speech patterns of Japanese learners of English.
5. A number of cloud-based speech recognition engines were evaluated to determine which is most appropriate for use with Japanese learners of English.
6. Spoken samples along with the corresponding automated scores are continually being collated and analyzed to investigate the accuracy of the automated scoring algorithm.
|
| 現在までの達成度 (区分) |
現在までの達成度 (区分)
2: おおむね順調に進展している
理由
We successfully completed a working version of the speech assessment system and implemented improvements in the system so it could be effectively integrated into the first year English curriculum during the 2018 academic year. At the time of this report, the system appears to be functioning as planned.
|
| 今後の研究の推進方策 |
User experience will be further improved to simplify the use of the system for both leaners and instructor.
The reliability of automated speech scoring algorithm will be improved, specifically with improved phoneme recognition, where a learner will produce part of a word correctly.
The validity of the speech scoring algorithm will be investigated via correlation checks of both human-rated speech tasks and computer-scored speech tasks.
|
| 次年度使用額が生じた理由 |
A nominal amount of 12,918 was not used in 2017. The programming costs were less than the original estimate. This amount will be used in 2018 to further develop the speech assessment system.
|
