Abstract | ||
---|---|---|
Pronunciation scoring is one important task for software designed to give feedback to students practicing a second language. English intonation can convey information about a speaker's nativeness, so previous studies have proposed using intonation-based models to score nonnative pronunciation. One past approach trained models for a set of pronunciation scores using ad hoc features derived from the frequency contour. We use prosodic theory to train models for categorical intonation units, inspired by work in modeling tone languages. These HMM-based models offer 0.398 correlation between automatic and listener scores on the ISLE nonnative speech corpus, compared to the 0.156 baseline correlation. |
Year | Venue | Keywords |
---|---|---|
2008 | INTERSPEECH 2008: 9TH ANNUAL CONFERENCE OF THE INTERNATIONAL SPEECH COMMUNICATION ASSOCIATION 2008, VOLS 1-5 | prosody, nonnative speech, pronunciation |
Field | DocType | Citations |
Computer science,Speech recognition | Conference | 7 |
PageRank | References | Authors |
0.82 | 5 | 2 |
Name | Order | Citations | PageRank |
---|---|---|---|
Joseph Tepperman | 1 | 73 | 8.59 |
Narayanan Shrikanth | 2 | 5558 | 439.23 |