In this paper, we investigate the network properties of the South Korean dialects. Among 153 words in “The Linguistic Atlas of Korea”, 33 words satisfying our criteria were surveyed. We set 138 basic administrative districts that make up South Korea as nodes and construct a language network by using a defined procedure. The network shows neither a small-world (small-world coefficient < 5) nor a scale-free (degree distribution forms like a unimodal Poisson) property in the general complex network. Also, the assortativity is seen to show transition from negative to positive at about the threshold value of 15. The number of Korean linguistic groups, which is the most import result in this study, is constantly four with s-curved Newman modularity in the threshold range 13 ∼ 22. The Korean dialect division by using a network modularity method is almost identical to the classification of traditional linguists (Central, Southwest, Southeast, and Jeju). Furthermore, subgroups can be obtained by performing a hierarchical dialect division. As a result, South Korean dialects are classified into seven minor groups, and some regions (e.g., Southern Gang-won) belong to groups different from those of the traditional classification method.

中文翻译：

---

韩国方言与传统方言分类对比的动态网络分析

在本文中，我们研究了韩国方言的网络特性。在《韩国语言地图集》的 153 个词中，有 33 个词符合我们的标准。我们将构成韩国的 138 个基本行政区设置为节点，并通过使用定义的程序构建语言网络。该网络在一般复杂网络中既不显示小世界（小世界系数 < 5）也不显示无标度（度分布形式如单峰泊松）属性。此外，在大约 15 的阈值处，分类性显示出从负到正的转变。 作为本研究中最重要的结果，韩语语言组的数量始终为 4，阈值为 s 曲线纽曼模块范围 13 ∼ 22。使用网络模块化方法进行的朝鲜语方言划分与传统语言学家（中部、西南、东南部和济州岛）的分类几乎相同。此外，可以通过执行分层方言划分来获得子群。因此，韩国方言被分为七个小群，一些地区（如江原南）属于与传统分类方法不同的群。