Pronunciation Error Detection Method based on Error Rule Clustering using a Decision Tree Export

@inproceedings{citeulike:326545, abstract = {We are developing a CALL system to train English pronunciation for Japanese native speakers. However, the precision of the error detection was not very high because the threshold for the detection was not optimum. To improve the detection accuracy, we propose a new method to optimize the thresholds of error detection. The proposed method makes several clusters of the pronunciation error rules, and the thresholds are determined for each cluster. An experiment was carried out to investigate the performance of the proposed method. As a result, about 90\% of detection rate was obtained, which is a remarkable improvement from the conventional method.}, author = {Ito, Akinori and Lim, Yen-Ling and Suzuki, Motoyuki and Makino, Shozo}, citeulike-article-id = {326545}, comment = {asr architecture similar to our tactlang system: error detection is performed by comparing canonical pronunciation to rule-generated likely mispronunciation. like our system, more accurate/lenient results are gained by smart thresholds. ---=note-separator=--- thresholds are generated by machine learning, clustering similar types of mistakes together. (although, ironically, it looks like the best clustering treats each mistake individually--that is, per-mistake thresholds instead of per-cluster thresholds)}, institution = {Tohoku University, Japan}, keywords = {asr, english, eurospeech, good, japanese, mispronunciation\_detection, modeling, pedagogy, sla}, month = {September}, organization = {Eurospeech}, posted-at = {2005-09-19 23:09:10}, priority = {0}, title = {Pronunciation Error Detection Method based on Error Rule Clustering using a Decision Tree}, year = {2005} }

TY - CONF ID - citeulike:326545 L3 - citeulike-article-id:326545 N2 - We are developing a CALL system to train English pronunciation for Japanese native speakers. However, the precision of the error detection was not very high because the threshold for the detection was not optimum. To improve the detection accuracy, we propose a new method to optimize the thresholds of error detection. The proposed method makes several clusters of the pronunciation error rules, and the thresholds are determined for each cluster. An experiment was carried out to investigate the performance of the proposed method. As a result, about 90% of detection rate was obtained, which is a remarkable improvement from the conventional method. OR - Eurospeech TI - Pronunciation Error Detection Method based on Error Rule Clustering using a Decision Tree KW - asr KW - english KW - eurospeech KW - good KW - japanese KW - mispronunciation_detection KW - modeling KW - pedagogy KW - sla N1 - asr architecture similar to our tactlang system: error detection is performed by comparing canonical pronunciation to rule-generated likely mispronunciation. like our system, more accurate/lenient results are gained by smart thresholds. N1 - thresholds are generated by machine learning, clustering similar types of mistakes together. (although, ironically, it looks like the best clustering treats each mistake individually--that is, per-mistake thresholds instead of per-cluster thresholds) AU - Ito, Akinori AU - Lim, Yen-Ling AU - Suzuki, Motoyuki AU - Makino, Shozo PY - 2005/09// ER -