| Presentation | 2003/3/20 Speech recognition of Japanese English using Japanese specific pronunciation habits Koichi OSAKI, Nobuaki MINEMATSU, Keikichi HIROSE, |
|---|---|
| PDF Download Page | PDF download Page Link |
| Abstract(in Japanese) | (See Japanese page) |
| Abstract(in English) | Today's globalization often requires Japanese people to talk in English, which is often seen in international meetings and so on. To apply speech recognition techniques into these situations successfully, improvement of non-native speech recognition has to be realised. The improvement can be achieved by enhancing acoustic models, pronunciation lexicon, language modeles, and/or decoding stragegies independently. In this paper, a research focus is put on the acoustic models. In our previous study, we proposed a technique to enhance MLLR-based adaptation with respect to pronunciation proficiency of the speaker, where proficiency-based phoneme regression trees were generated and used in the adaptation. But this method had several drawbacks. It cannot be applied to speakers with lower proficiency or give any solutions to a problem that the commonly-used top-down state-tying may introduce inadequate phonetic structure into the model set. In this paper, we firstly analyze Japanese English and propose a technique of combining Japanese English acoustic models with Japanese ones. Next, we investigate a new state-tyine technique which considers spellings of vowels as well as left and right phoneme context. |
| Keyword(in Japanese) | (See Japanese page) |
| Keyword(in English) | non-native speech recognition / Japanese English / pronunciation proficiency / multi-path models / spell |
| Paper # | SP2002-180 |
| Date of Issue |
| Conference Information | |
| Committee | SP |
|---|---|
| Conference Date | 2003/3/20(1days) |
| Place (in Japanese) | (See Japanese page) |
| Place (in English) | |
| Topics (in Japanese) | (See Japanese page) |
| Topics (in English) | |
| Chair | |
| Vice Chair | |
| Secretary | |
| Assistant | |
| Paper Information | |
| Registration To | Speech (SP) |
|---|---|
| Language | JPN |
| Title (in Japanese) | (See Japanese page) |
| Sub Title (in Japanese) | (See Japanese page) |
| Title (in English) | Speech recognition of Japanese English using Japanese specific pronunciation habits |
| Sub Title (in English) | |
| Keyword(1) | non-native speech recognition |
| Keyword(2) | Japanese English |
| Keyword(3) | pronunciation proficiency |
| Keyword(4) | multi-path models |
| Keyword(5) | spell |
| 1st Author's Name | Koichi OSAKI |
| 1st Author's Affiliation | Graduate School of Information Science and Technology, University of Tokyo() |
| 2nd Author's Name | Nobuaki MINEMATSU |
| 2nd Author's Affiliation | Graduate School of Information Science and Technology, University of Tokyo |
| 3rd Author's Name | Keikichi HIROSE |
| 3rd Author's Affiliation | Graduate School of Frontier Sciences, University of Tokyo |
| Date | 2003/3/20 |
| Paper # | SP2002-180 |
| Volume (vol) | vol.102 |
| Number (no) | 749 |
| Page | pp.pp.- |
| #Pages | 6 |
| Date of Issue | |