| Presentation | 2008-12-10 Verification of Speech Recognition Results Based on the Utterance Classification Using Conditional Random Fields Kenko OTA, Terumasa EHARA, |
|---|---|
| PDF Download Page |  PDF download Page Link |
| Abstract(in Japanese) | (See Japanese page) |
| Abstract(in English) | A dialogue strategy for disambiguating user utterances is discussed in this study. In case where the systems cannot understand user utterances because of speech recognition errors, users are annoyed with the iterative confirmation. So, a method for automatic verification and modification of the speech recognition results is proposed. Many supposed utterances clustered in accordance with content similarity are prepared for estimating user intention. Then, a speech recognition result is classified into a cluster using conditional random fields. The system decides whether rerecognizing a part of the utterances supposed to be mistaken or generating a confirmation dialogue in order to be respoken based on an utterance classification result. Estimating and modifying speech recognition errors by the system itself based on a user utterance, users may get relief from iterative confirmation. From the evaluation results, it is possible that the proposed method can classify speech recognition results into clusters robustly and the proposed method can detect and modify the speech recognition errors automatically based on the utterance classification results. |
| Keyword(in Japanese) | (See Japanese page) |
| Keyword(in English) | Spoken dialogue system / Natural language processing / Conditional random field / Utterance intention / Speech recognition verification |
| Paper # | NLC2008-68,SP2008-123 |
| Date of Issue |
| Conference Information | |
| Committee | NLC |
|---|---|
| Conference Date | 2008/12/2(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 | Natural Language Understanding and Models of Communication (NLC) |
|---|---|
| Language | JPN |
| Title (in Japanese) | (See Japanese page) |
| Sub Title (in Japanese) | (See Japanese page) |
| Title (in English) | Verification of Speech Recognition Results Based on the Utterance Classification Using Conditional Random Fields |
| Sub Title (in English) | |
| Keyword(1) | Spoken dialogue system |
| Keyword(2) | Natural language processing |
| Keyword(3) | Conditional random field |
| Keyword(4) | Utterance intention |
| Keyword(5) | Speech recognition verification |
| 1st Author's Name | Kenko OTA |
| 1st Author's Affiliation | Tokyo University Science, Suwa Department of System Engineering, Faculty of Electronic System Engineering() |
| 2nd Author's Name | Terumasa EHARA |
| 2nd Author's Affiliation | Tokyo University Science, Suwa Department of System Engineering, Faculty of Electronic System Engineering |
| Date | 2008-12-10 |
| Paper # | NLC2008-68,SP2008-123 |
| Volume (vol) | vol.108 |
| Number (no) | 337 |
| Page | pp.pp.- |
| #Pages | 6 |
| Date of Issue | |