The impact of language transfer on native speaker recognition of native and non-native speech

Abstract

There are many identifiable acoustic differences between native and non-native English, including patterns of prosody (Ramirez Verdugo, 2005), vowel quality (Flege et al., 1997) and extent of reduction in speech (Baker et al., 2011), many of which arise due to transfer from the native language (L1 transfer). This study examines the effect that differences generated through language transfer have upon the recognition of the Mandarin-accented English compared to American English by native speakers of American English. Listeners were presented with a set of 40 different TIMIT sentences (Garofolo et al., 1993) spoken by 5 native speakers of American English and 5 native Mandarin speakers, and were asked to transcribe what they heard. These transcriptions will be evaluated for accuracy, with errors identified and classified according to the type(s) of acoustic deviations from native English speech attested in any misheard words, as well as whether these errors are typical of Mandarin-speaking learners of English. Results will be used to evaluate the relative importance of these differences between native and non-native English speech in recognition and to identify potentially problematic speech patterns for use with multi-dialectal systems of automatic speech recognition.