Outliers or Wrong Classifications? On Dialect Typicality and Transitional Dialects
Ho Wang Matthew Sung
Leiden University
Since dialect areas do not have abrupt borders (c.f. Meyer 1877, Paris 1888, Jaberg 1908, Viereck 1992), one could speak of ‘focal’ or ‘core’ and ‘transitional’ areas (Taeldeman and Hinskens 2013: 131). Transitional dialects are defined as “dialects possessing only features present in one or another neighbouring area” (Ivic 1962: 43). The descriptions and analyses of transitional dialects mentioned above are however mostly speculative or based on single-feature analyses. This means that it is still unclear whether these generalisations can be directly applied on an aggregate analysis (multiple features in one single analysis, Nerbonne 2010). Afterall, dialects do not only possess one feature, but numerous features.
In this presentation, I aim to explore how dialects transition from one area to another on the aggregate level by looking at the Guangfu dialects from the Yue-speaking area. The dataset consists of multi-aligned IPA transcriptions of 130 words in 104 Yue and Southern Pinghua dialects (Sung, Prokic and Chen 2024).
In order to assess the transitionality of dialects, this presentation will introduce a novel data-driven method called Dialect Typicality Analysis. This method involves the following steps: 1) identification of dialect areas, 2) features extraction and 3) typicality measurement. Firstly, Guangfu dialects are identified by cluster analysis (Ward’s method, Ward 1963), based on the dialect distances calculated using Relative Distance Value (Goebl 2018), as illustrated in Sung and Prokic (2023, submitted). Next, features closely associated with Guangfu dialects are extracted using Normalised Pointwise Mutual Information (Sung and Prokic 2024). This step gives a list of features ranked by the strength of association to the dialect group. Finally, a typicality analysis is conducted using a typicality ‘decay’ plot, which shows the accumulated typicality per 10 features of each dialect based on the feature list in the previous step.
The result of the typicality analysis shows that transitional dialects can possess highly typical features of one dialect group, while the majority of its features pattern with another dialect group (suggesting borrowing or contact phenomenon). One of these dialects is the Yangshan dialect. On the decay plot, dialects from Guangfu (Guangzhou, Foshan, Panyu, Shunde, Nanhai) and Goulou (Yangshan, Huaiji, Sihui, Guangning) dialect groups (identified by the cluster analysis) are reflected by the two bundles of lines on the decay plot, which represent the difference in the (Guangfu) typicality patterns as we increase the features considered. The Yangshan dialect, however, shows a different pattern from these two dialect groups. Despite it possesses the top 10 features which Guangfu dialects have (the two features mentioned above, found in multiple words), as we increase the number of features, the Guangfu typicality of the Yangshan dialect gradually drops and bundles with the rest of the Goulou dialects. This explains why the Yangshan dialect is classified as a Goulou dialect, while possessing the most exclusive Guangfu features. This novel method can be used in complement with cluster analysis in order to understand the gradual relationship between dialect groups, which hard clustering cannot offer.
References:
Goebl, H. (2018). Dialectometry.
Ivić, P. (1962) On the Structure of Dialectal Differentiation.
Jaberg, K. (1908). Sprachgeographie. Beitrag zum Verständnis des Atlas linguistique de la France.
Meyer, P. (1877). PÉRIODIQUES: Archives des missions scientifiques et littéraires. 3e série, t. III, 2e livraison.
Nerbonne, J. (2010). Mapping aggregate variation.
Paris, G. (1888). Les parlers de France.
Sung, H. W. M. & Prokić, J. (2023) What are Guangfu dialects?. [Presentation at the 27th International Conference on Yue Dialects, December 2023]
Sung, H. W. M. & Prokić, J. (2024). Detecting Dialect Features Using Normalised Pointwise Information.
Sung, H. W. M. & Prokić, J. (submitted). Relative Chronology of Dialectal Phonetic Features. In Relative chronology in historical linguistics, ed. by F. Wandl, T. Olander & J.-M. List. Language Science Press (in preperation).
Sung, H. W. M., Prokic, J. and Chen C. (2024). A New Dataset for Tonal and Segmental Dialectometry from the Yue- and Pinghua-Speaking Area.
Taeldeman, J., & Hinskens, F. L. M. P. (2013). The classification of the dialects of Dutch.
Viereck, W. (1992). Prince Louis-Lucien Bonaparte and English Dialectology.
Ward Jr, J. H. (1963). Hierarchical grouping to optimize an objective function.