Bipartite spectral graph partitioning for clustering dialect varieties and detecting their linguistic features

Abstract

In this study we use bipartite spectral graph partitioning to simultaneously cluster varieties and identify their most distinctive linguistic features in Dutch dialect data.While clustering geographical varieties with respect to their features, e.g. pronunciation, is not new, the simultaneous identification of the features which give rise to the geographical clustering presents novel opportunities in dialectometry. Earlier methods aggregated sound differences and clustered on the basis of aggregate differences. The determination of the significant features which co-vary with cluster membership was carried out on a post hoc basis. Bipartite spectral graph clustering simultaneously seeks groups of individual features which are strongly associated, even while seeking groups of sites which share subsets of these same features. We show that the application of this method results in clear and sensible geographical groupings and discuss and analyze the importance of the concomitant features.Key words: Bipartite spectral graph partitioning, Clustering, Sound correspondences, Dialectometry, Dialectology, Language variation $ This paper is an extended version of the study 'Bipartite spectral graph partitioning to co-cluster varieties and sound correspondences in dialectology' by Martijn Wieling and John Nerbonne [1]

A multivariate spatial analysis of vowel formants in American English

“…see Prokić & Nerbonne, 2008;Nerbonne & Heeringa, 2009;Wieling & Nerbonne, 2010;Grieve et al, 2011). In particular, Ward's method for hierarchical clustering (Ward, 1963) was used because it tends to identify the clearest dialect regions and because it is one of the most common methods for hierarchical clustering in dialectometry.…”

A comparison of statistical methods for the aggregation of regional linguistic variation

“…Séguy initiated the field of dialectometry to overcome the limitations of the isogloss method, but the analysis of individual linguistic variables and the identification of subsets of variables that exhibit similar patterns are still worthwhile. Most dialectometry analyses, however, focus on the linguistic distance matrix from which information about the patterns exhibited by the individual linguistic variables cannot be extracted directly (although for dialectometry research that addresses some of these issues see Shackleton, 2005Shackleton, , 2007Nerbonne, 2006;Rumpf et al, 2009Rumpf et al, , 2010Wieling and Nerbonne, 2011). These limitations with the standard approach to dialectometry led to the development of an alternative statistical approach to the analysis of regional linguistic variation known as a multivariate spatial analysis (Grieve, 2009;Grieve et al, forthcoming).…”

“…While the multidimensional scaling identifies continuous patterns of aggregated regional linguistic variation, a cluster analysis can be used to produce a discrete classification of the locations, which can then be mapped in order to identify absolute patterns of aggregated regional linguistic variation. In this analysis, the linguistic distance matrix was subjected to a hierarchical cluster analysis (Shackleton, 2005(Shackleton, , 2007Goebl, 2007;Prokic & Nerbonne, 2008;Wieling & Nerbonne, 2010). A hierarchical cluster analysis identifies clusters of similar objects in a distance matrix by initially assigning each observation to its own cluster and by then repeatedly combining the two most similar clusters to form larger and larger clusters until all of the objects have been combined to form one large cluster.…”

Using Gabmap

“…An earlier method based loosely onWieling and Nerbonne (2011a) is also available for categorical data.…”