Sixth International Conference on Spoken Language Processing
Speech utterances in a given language are known to be less intelligible to second-language (L2) listeners than to native listeners. This study is aimed at quantifying the reduction of speech intelligibility due to non-nativeness of the listener. Two types of experiments were carried out: Speech Reception Threshold (SRT) experiments for obtaining a quantitative measure of speech intelligibility, and Letter Guessing Procedure (LGP) experiments to assess to influence of linguistic factors of non-nativeness on speech intelligibility separately. The effects of non-nativeness of listeners were found to be in the range of 1.4 to 4.4 decibel, in terms of speech-to-noise ratio that gives 50% sentence intelligibility. The magnitude of this effect depends on linguistic experience. There is a good correlation between intelligibility and linguistic entropy. To highly proficient L2 listeners, a non-native accent in L2 (similar to their own accent) reduces intelligibility. Less proficient L2 listeners do benefit from such a non-native accent in L2 speech. This indicates a na´ve phonetic discrimination.
Bibliographic reference. Wijngaarden, Sander J. van / Steeneken, Herman J.M. (2000): "The intelligibility of German and English speech to Dutch listeners", In ICSLP-2000, vol.3, 929-932.