| Presentation | 2010-12-21 Evaluation of Successive Rapid Hypothesis Determination Algorithm for Continuous Word Recognition Hiroyuki OHNO, Hiroshi KOJIMA, Yoshihiko NANKAKU, Akinobu LEE, Keiichi TOKUDA, |
|---|---|
| PDF Download Page |  PDF download Page Link |
| Abstract(in Japanese) | (See Japanese page) |
| Abstract(in English) | Minimizing response delay of speech recognition system and giving rapid feed backs are important properties for an intuitive, easy-to-use speech interfaces. Many studies has been conducted to improve the response delay, such as making progressive outputs while recognition process "after" the words are half-determined in the context. In order to achieve higher speed input responses, we have proposed an algorithm to determine the most likely hypothesis "before" the utterance ends. The method has been examined for isolated word recognition, and this paper extends it for continuous word recognition. Experimental evaluations were performed for tasks of various vocabulary size. The result at a small vocabulary task with 14 words has shown that our proposed algorithm can determine each word for about 0.053 second prior to the actual end of speech on average, without any degradation of recognition accuracy. Another result on a station names recognition task with vocabulary size of 8738 has shown that our proposed algorithm can determine each word for about 0.48 second on average after the actual end of speech. The comparison results on various acoustic models are also reported. |
| Keyword(in Japanese) | (See Japanese page) |
| Keyword(in English) | Speech recognition / search algorithm / progressive output / tree lexicon / confidence measure |
| Paper # | NLC2010-21,SP2010-94 |
| Date of Issue |
| Conference Information | |
| Committee | NLC |
|---|---|
| Conference Date | 2010/12/13(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) | Evaluation of Successive Rapid Hypothesis Determination Algorithm for Continuous Word Recognition |
| Sub Title (in English) | |
| Keyword(1) | Speech recognition |
| Keyword(2) | search algorithm |
| Keyword(3) | progressive output |
| Keyword(4) | tree lexicon |
| Keyword(5) | confidence measure |
| 1st Author's Name | Hiroyuki OHNO |
| 1st Author's Affiliation | Department of Computer Science and Engineering, Nagoya Institute of Technology() |
| 2nd Author's Name | Hiroshi KOJIMA |
| 2nd Author's Affiliation | Department of Computer Science and Engineering, Nagoya Institute of Technology:(Present Office)Hitachi Solutions, Ltd. |
| 3rd Author's Name | Yoshihiko NANKAKU |
| 3rd Author's Affiliation | Department of Computer Science and Engineering, Nagoya Institute of Technology |
| 4th Author's Name | Akinobu LEE |
| 4th Author's Affiliation | Department of Computer Science and Engineering, Nagoya Institute of Technology |
| 5th Author's Name | Keiichi TOKUDA |
| 5th Author's Affiliation | Department of Computer Science and Engineering, Nagoya Institute of Technology |
| Date | 2010-12-21 |
| Paper # | NLC2010-21,SP2010-94 |
| Volume (vol) | vol.110 |
| Number (no) | 356 |
| Page | pp.pp.- |
| #Pages | 6 |
| Date of Issue | |