CHAOJU TANG

Leiden University, THE NETHERLANDS

Predicting mutual intelligibility among Chinese dialects from subjective and objective linguistic similarity measures

Dialectología

We are interested in determining mutual intelligibility of Chinese (Sinitic) dialects. Theoretical and quantitative work has been done on a set of 17 Chinese dialects by Cheng (1997), who computed a theoretical construct he called mutual intelligibility from lexical, phonological and syntactic properties. We aimed to validate the theoretical construct using experimental judgment data collected from naive language users.

We targeted the following 15 dialects (a proper subset from Cheng 1997): Beijing, Chengdu, Jinan, Xi’an, Taiyuan, Hankou (Mandarin dialects), Suzhou, Wenzhou,Yangzhou (Wu dialects), Nanchang (Gan dialect), Meixian, Chaozhou (Hakka dialects), Xiamen, Fuzhou (Min dialects), and Changsha (Xiang dialect). We manipulated readings of the fable “The North Wind and the Sun such that all speakers sounded like males (an acoustic gender transformation was carried out on the female speakers), all had roughly the same articulation rate and speech-pause ratio, and the same mean pitch. Each fable was resynthesized in two melodic versions, i.e., one with the original pitch intervals kept intact, and one with a constant pitch equal to the mean pitch of the fragment with melody (and the same as all other fragments). The 30 readings of the fable were presented to 15 groups of dialect listeners in China (same 15 dialects as those of the speakers, with 12 male and 12 female listeners in each dialect group). Listeners rated all materials twice: first, they estimated how well they believed they could understand the speaker between ‘0’ for ‘I do not understand a single word’ and ‘10’ for ‘I understand the speaker perfectly’. Second, listeners rated the similarity between their own dialect and that of each speaker in the recording, where ‘0’ meant ‘No similarity at all’ against ’10’ meaning ‘This dialect is exactly the same as my own dialect’. In all 21,600 judgments were collected.

We statistically regressed Cheng’s objective computational distance measures against our judgment data in order to determine how well judged intelligibility and similarity can be predicted by the objective measurements. The results show that the objective measures afford fairly good prediction of the judgment scores, and that both sets of measures correspond reasonably well with traditional genealogical trees published by Chinese dialectologists. Judged intelligibility and judged similarity were strongly correlated except in the case of Beijing dialect. Since this dialect is virtually the same as the standard language, it was rated as highly intelligible by most dialect listeners, and yet very dissimilar to the native dialect.




Asociación de Jóvenes Lingüistas
ajl2006@gmail.com