| Project/Area Number |
18H04107
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (A)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Review Section |
Medium-sized Section 61:Human informatics and related fields
|
|---|
| Research Institution | The University of Tokyo |
|---|
Principal Investigator |
Minematsu Nobuaki 東京大学, 大学院工学系研究科(工学部), 教授 (90273333)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
牧野 武彦 中央大学, 経済学部, 教授 (00269482)
山内 豊 創価大学, 教育学部, 教授 (30306245)
齋藤 大輔 東京大学, 大学院工学系研究科(工学部), 准教授 (40615150)
|
|---|
| Project Period (FY) |
2018-04-01 – 2022-03-31
|
| Project Status |
Completed (Fiscal Year 2022)
|
| Budget Amount *help |
¥40,560,000 (Direct Cost: ¥31,200,000、Indirect Cost: ¥9,360,000)
Fiscal Year 2021: ¥10,270,000 (Direct Cost: ¥7,900,000、Indirect Cost: ¥2,370,000)
Fiscal Year 2020: ¥9,490,000 (Direct Cost: ¥7,300,000、Indirect Cost: ¥2,190,000)
Fiscal Year 2019: ¥10,010,000 (Direct Cost: ¥7,700,000、Indirect Cost: ¥2,310,000)
Fiscal Year 2018: ¥10,790,000 (Direct Cost: ¥8,300,000、Indirect Cost: ¥2,490,000)
|
|---|
| Keywords | 外国語教育 / シャドーイング / 瞬時的了解度 / 定量的評価 / ポステリオグラム / 深層学習 / 音声認識 / 発音評価 / 瞬時的明瞭度 / 聴解モデリング / 瞬時的了解性 / 可解性と了解性 / 定量的計測 / 逆シャドーイング / 可解性 / 事後確率DTW / comprehensibility / 音素事後確率 |
|---|
| Outline of Final Research Achievements |
In language education, shadowing is a training technique to enhance listening skills, where learners are asked to repeat a given model speech while listening to it. Since shadowing can be viewed as a technique to have listeners report what they just heard while listening, we asked native listeners or raters to shadow learners' speech. Here, raters' shadowing breakdowns were treated as listening breakdowns, and they were used to quantify easiness or difficulty of identifying words (intelligibility) in learners' speech. This quantification was made in sequence, and sequential intelligibility was called as instantaneous intelligibility. After collecting 750 Japanese accented English utterances, we asked three raters' with different language backgrounds to shadow all the utterances, and their listening breakdowns were modeled to built an automatic predictor of sequential and instantaneous intelligibility for any new learner's utterance.
|
| Academic Significance and Societal Importance of the Research Achievements |
音声認識や音声合成に代表される音声技術を用いた外国語音声教育支援は，その多くが「発音」教育支援であり，「聴取」教育支援は極めて少ない。これは「聴取」している様子は音響的に計測できないため，そもそも音声技術の対象外として扱われてきたことに因る。本研究では，これまで研究対象とされてこなかった「聴取の崩れ」の音響的計測を可能にした。そして，数百名の日本人英語音声を評価者にシャドーさせることで，評価者の聴き淀みをモデル化し，新たな音声に対してどこで聴き淀むのかを予測する技術を構築した。このように本研究は学術的意義が高く，また，外国語音声（聴取）教育という，実用価値の高い分野に応用することができた。
|
