Wenda Chen Institute for Infocomm Research, A*STAR, Singapore and - PowerPoint PPT Presentation

Wenda Chen

 Institute for Infocomm Research, A*STAR, Singapore and Nanyang Technological University, Singapore  NTU project: Computer Assisted Language Learning System in Singapore English  I 2 R project: Statistical analysis of the Corpus based pronunciation variation rule for pronunciation error detection and feedback in Singapore English

 Between training and testing data set

Buckeye corpus

 Dialect and mispronunciation detection  E{C=>F|b=K,d=M} estimation  F=g(C), estimate E{R|Test} based on E{R|Sg} and E{R|Am}  Statistical test, p value, binomial test, t test (degree of freedom: no. of testing phones), multinomial distribution of phone errors in SgE data, AmE data and test data  MAP framework for phoneme error event detection and finding the optimum threshold combination  Vector subspace decomposition and representation of original AmE into SgE phone set