Interspeech'2005 - Eurospeech

Lisbon, Portugal
September 4-8, 2005

Automatic Detection of Frequent Pronunciation Errors Made by L2-Learners

Khiet P. Truong (1), Ambra Neri (2), Febe de Wet (3), Catia Cucchiarini (2), Helmer Strik (2)

(1) TNO Human Factors, The Netherlands; (2) Radboud Universiteit Nijmegen, The Netherlands; (3) University of Stellenbosch, South Africa

In this paper, we present an acoustic-phonetic approach to automatic pronunciation error detection. Classifiers using techniques such as Linear Discriminant Analysis and Decision Trees were developed for three sounds that are frequently pronounced incorrectly by L2-learners of Dutch: /A/, /Y/ and /x/. This paper will focus mainly on the problems with the latter phoneme. The acoustic properties of these pronunciation errors were examined so as to define a number of discriminative acoustic features to be used to train and test the classifiers. Experiments showed that the classifiers are able to discriminate correct sounds from incorrect sounds in both native and non-native speech, and therefore can be used to detect pronunciation errors in non-native speech.

Full Paper

Bibliographic reference.  Truong, Khiet P. / Neri, Ambra / Wet, Febe de / Cucchiarini, Catia / Strik, Helmer (2005): "Automatic detection of frequent pronunciation errors made by L2-learners", In INTERSPEECH-2005, 1345-1348.