2021 Fiscal Year Final Research Report
Pronunciation and accent modeling for multi-dialect speech synthesis
| Project/Area Number |
18K18100
|
|---|
| Research Category |
Grant-in-Aid for Early-Career Scientists
|
| Allocation Type | Multi-year Fund |
|---|
| Review Section |
Basic Section 61030:Intelligent informatics-related
|
|---|
| Research Institution | The University of Tokyo |
|---|
Principal Investigator |
|
|---|
| Project Period (FY) |
2018-04-01 – 2022-03-31
|
| Keywords | 音声合成 / 方言 / 韻律 |
|---|
| Outline of Final Research Achievements |
The purpose of this research is to artificially synthesize speech in any Japanese dialect. To achieve this goal, we have developed (1) a method that enables robust speech synthesis from noisy recorded speech, (2) a speech synthesis method that controls accents using geographic information of dialects, (3) a method for acquiring linguistic units for constructing accents without linguistic knowledge, (4) a method for acquiring dialectal accents without linguistic knowledge, (5) a method for realizing dialectal speech synthesis without linguistic knowledge, and (6) the release of a free speech database to realize dialectal speech synthesis.
|
| Free Research Field |
知能情報学
|
| Academic Significance and Societal Importance of the Research Achievements |
本研究は，あらゆる日本語方言の音声を人工的に合成することを目的とする．消滅の危機にある日本語方言について，その特性を計算機的に保存することは，音声言語文化の保存からコンテンツ制作まで幅広い範囲に有用である．本研究はこれに向け，方言の知識なしに方言音声を合成可能な方法について多角的に取り組み，さらに，一般に利用可能な方言データベースを整備した．
|
