OJADを支える音声合成技術 音声を合成しない音声合成技術の使い方 Speech synthesis technology that supports OJAD An example of applying speech synthesis techniques not to synthesizing speech

この論文にアクセスする

この論文をさがす

著者

抄録

日本語の韻律教育を支援すべく,自然言語処理技術,音声言語処理技術を用いたオンラインアクセント辞書 (Online Japanese Accent Dictionary, OJAD) [1] を構築,運用している。日本語は前後のコンテキストによって単語のアクセントが頻繁に変化する特徴を有するが,アクセント変形に十分対応した日本語教育史上初の教材として,世界中の教育現場で利用されるに至っている。またこれまで,約 4 時間に渡る OJAD 講習会を,国内 10 都市,海外 17 都市で開催しており,どの講習会も好評を博している。OJAD 開発を技術的観点から見ると,アクセント句境界推定,アクセント核位置推定,Fo パターン生成など,音声合成の裏方として機能していた技術を表舞台に出しているに過ぎない。これは音声合成技術の一部を,音声を合成する目的以外に応用している例として考えることができる。本稿では,音声合成技術の応用可能性を考える一つの例として OJAD 開発・運用を捉え,検討する。To support Japanese prosody instruction, the Online Japanese Accent Dictionary (OJAD) [1] has been developed by using NLP and SLP techniques and it is maintained by our laboratory. Japanese is a very unique language in that word accent often changes due to its context. The OJAD was introduced to the Japanese language education community as the first educational system that can handle context-based word accent changes very well and it is actively used by teachers and learners internationally. So far, 4-hour OJAD tutorials have been held at 10 domestic cities and 17 international cities and each tutorial was welcomed to Japanese teachers there. If we discuss development of the OJAD from a technical point of view, the OJAD uses several internal modules of Japanese speech synthesis, such as estimation of accent phrase boundaries and accent nucleus positions, and Fo pattern generation. It is interesting that the OJAD uses these techniques not for synthesizing speech. In this report, by regarding development of the OJAD as one example of using speech synthesis techniques not to synthesizing speech, we discuss new possibility of applying these techniques to new domains.

To support Japanese prosody instruction, the Online Japanese Accent Dictionary (OJAD) [1] has been developed by using NLP and SLP techniques and it is maintained by our laboratory. Japanese is a very unique language in that word accent often changes due to its context. The OJAD was introduced to the Japanese language education community as the first educational system that can handle context-based word accent changes very well and it is actively used by teachers and learners internationally. So far, 4-hour OJAD tutorials have been held at 10 domestic cities and 17 international cities and each tutorial was welcomed to Japanese teachers there. If we discuss development of the OJAD from a technical point of view, the OJAD uses several internal modules of Japanese speech synthesis, such as estimation of accent phrase boundaries and accent nucleus positions, and Fo pattern generation. It is interesting that the OJAD uses these techniques not for synthesizing speech. In this report, by regarding development of the OJAD as one example of using speech synthesis techniques not to synthesizing speech, we discuss new possibility of applying these techniques to new domains.

収録刊行物

  • 研究報告音声言語情報処理(SLP)

    研究報告音声言語情報処理(SLP) 2013-SLP-99(21), 1-6, 2013-12-12

    一般社団法人情報処理学会

各種コード

  • NII論文ID(NAID)
    110009646544
  • NII書誌ID(NCID)
    AN10442647
  • 本文言語コード
    JPN
  • 資料種別
    Technical Report
  • ISSN
    0913-5685
  • NDL 記事登録ID
    024342465
  • NDL 請求記号
    Z16-940
  • データ提供元
    NDL  NII-ELS  IPSJ 
ページトップへ