Research on communication disconnection prevention technology conducive to dialogue with conversation robot
| Project/Area Number |
16K16134
|
|---|
| Research Category |
Grant-in-Aid for Young Scientists (B)
|
| Allocation Type | Multi-year Fund |
|---|
| Research Field |
Intelligent robotics
|
|---|
| Research Institution | Kagawa National College of Technology |
|---|
Principal Investigator |
SASAYAMA MANABU 香川高等専門学校, 情報工学科, 講師 (60508232)
|
|---|
| Project Period (FY) |
2016-04-01 – 2019-03-31
|
| Project Status |
Completed (Fiscal Year 2018)
|
| Budget Amount *help |
¥1,040,000 (Direct Cost: ¥800,000、Indirect Cost: ¥240,000)
Fiscal Year 2018: ¥260,000 (Direct Cost: ¥200,000、Indirect Cost: ¥60,000)
Fiscal Year 2017: ¥130,000 (Direct Cost: ¥100,000、Indirect Cost: ¥30,000)
Fiscal Year 2016: ¥650,000 (Direct Cost: ¥500,000、Indirect Cost: ¥150,000)
|
|---|
| Keywords | 問い返し / 聞き間違い / 言い間違い / コミュニケーション / 言い間違いの検出 / 聞き間違いの検出 / 言い間違いコーパス / 聞き間違いコーパス / 音位転換 / 字幕情報 / 対話アプリ / 対話API / 聞き返し / 対話コーパス / 音素 / 音声間距離 / 情報工学 / 感性情報学 |
|---|
| Outline of Final Research Achievements |
We collected mishearing utterances from the web and constructed a mishearing corpus.In the detection method of mishearing, first, the readings of all the original utterances are input and converted to phonemes.Next, we output a candidate for misinterpretation from the phoneme sequence.Finally, the misplacement is detected by calculating the degree of similarity to the mishearing utterance.As a result of conducting the detection experiment, we got a detection accuracy of 83% of misplaced parts.We collected speech errors from books and the web, and built a speech error corpus.We created a model by performing deep learning using CNN and LSTM.We conducted a detection experiment of speech errors, we obtained about 70% detection accuracy in the case of LSTM.
|
| Academic Significance and Societal Importance of the Research Achievements |
聞き間違いを含む対話を収集して分析することで人間が聞き間違いやすい音素や拍の変動などが明らかになった．このことから，コミュニケーションロボットが人間の問い返しに対応できるようになると共にコミュニケーションロボットが発話する単語や表現を聞き取りやすい単語や表現に変更できる．また，言い間違いを含む対話の分析からコミュニケーションロボットが人間の言い間違いを認識できる．これらの成果が現在のコミュニケーションロボットと人間のコミュニケーションの断絶を防止でき，ストレスフリーな対話が実現できる．
|
Report
(4 results)
Research Products
(8 results)
