Abstract | ||
---|---|---|
Correctly reading letter-sounds is an essential first step towards reading words and sentences. Pronunciation assessment of letter-sounds is an important component of preliterate children's education, and automating this process can have several advantages. We propose a method to automatically verify the pronunciations of a letter-sound task administered to kindergarteners and first graders in realistic noisy classrooms. We compare different acoustic models, decoding grammars, and dictionaries to help differentiate between acceptable and unacceptable pronunciations. Our final system achieved 88.0 percent agreement (0.702 kappa agreement with expert human evaluators, who agree themselves 94.9 percent of the time (0.886 kappa agreement). |
Year | Venue | Keywords |
---|---|---|
2008 | INTERSPEECH 2008: 9TH ANNUAL CONFERENCE OF THE INTERNATIONAL SPEECH COMMUNICATION ASSOCIATION 2008, VOLS 1-5 | children's speech, pronunciation verification, automatic reading assessment, letter-sounds |
Field | DocType | Citations |
Pronunciation,Computer science,Speech recognition | Conference | 6 |
PageRank | References | Authors |
0.53 | 5 | 5 |
Name | Order | Citations | PageRank |
---|---|---|---|
Matthew P. Black | 1 | 192 | 13.67 |
Joseph Tepperman | 2 | 73 | 8.59 |
Abe Kazemzadeh | 3 | 957 | 52.95 |
Sungbok Lee | 4 | 1394 | 84.13 |
Narayanan Shrikanth | 5 | 5558 | 439.23 |