Assessment of non-native speech using vowel space characteristics

Abstract

In this paper, we propose the idea of using the characteristics of a speaker's vowel space for automated assessment of second language (L2) proficiency. Specifically, we adpot features that were shown in previous studies to be good indicators of native speaker intelligibility and clarity and apply them to L2 speech from non-native speakers. The features focus on three peripheral vowels (IY, AA, and OW) and measure a speaker's coverage of the vowel space. A pilot study and a large-scale corpus study involving read speech produced by native and non-native speakers were conducted in which the vowel space features were rank correlated with pronunciation scores provided by human listeners for the non-native speech and an assumed higher score for the native speech. The results of the studies show that several of the features achieve moderately high correlations with the pronunciation scores, supporting their usefulness for automated assessment of non-native speech. The feature with the best performance in the largescale study was the F2 − F1 distance for IY, which achieved a correlation of 0.78 with pronunciation proficiency scores.

Static measurements of vowel formant frequencies and bandwidths: A review

“…Several studies have shown that VSA is correlated with speech intelligibility, with larger values of VSA being associated with higher levels of intelligibility in normal native speech (de Boer, 2009; Neel, 2008; Smiljanić & Bradlow, 2009), normal non-native speech (Chen, Evanini, & Sun, 2010), and disordered speech (de Bruijn et al, 2009; Higgins & Hodge, 2002; Turner, Tjaden, & Weismer, 1995). VSA is only one factor that accounts for the intelligibility of speech, and its importance in any given speaker or group of speakers depends on other factors, such as the availability of different types of acoustic cues.…”

Gaussian Mixture Modeling of vowel durations for automated assessment of non-native speech

“…Intuitively, the speech dissimilarity between native and non-native speakers is much larger than among non-native speakers themselves. A study of vowel space characteristics between native and non-native speakers [8] provides support for this intuition. So it may be meaningful to use non-native speaker data to assess non-native speech.…”

Vowel Characteristics in the Assessment of L2 English Pronunciation

“…We have presented the results of a study examining the utility of spectral features of vowels in the automated assessment of non-native English speech. Previous studies ( [9], [10]) have used vowel space size (or vowel dispersion) as a measure of accuracy in L2 pronunciation. However, the theoretical basis for this assumption remains unclear.…”

“…Recent studies that have examined the use of formant characteristics in the automated assessment of speech intelligibility (e.g. [9]; [10]) have established a link between proficiency and vowel formant characteristics. It is regrettable, however, that no serious attempt has been made in the vast majority of these studies to normalise speaker differences in vocal tract sizes.…”