外向型融媒汉语学习词典与大语言模型释义对比研究——以汉语方言词为例

A Comparative Study on Definitions between Outward-Oriented Media-Integrated Chinese Learning Dictionaries and Large Language Models—A Case Study of Chinese Dialect Words

摘要: 外向型汉语学习词典是国际中文教育的重要工具，但其方言词的释义长期缺乏统一标准。以《JUZI汉语》为代表的融媒词典正在尝试突破传统编纂模式，而大语言模型作为新兴的汉语检索学习途径，其语言处理能力同样受到关注。两者均为学习者可通过网络直接获取的语言资源，但二者在具体方言词的释义差异尚缺乏系统的实证比较。本研究以《JUZI汉语》为融媒词典案例，以DeepSeek和豆包为两个大语言模型代表，将三者进行对比：首先对《JUZI汉语》收录的方言词进行难度标注，继而再依据语义透明度理论，在每个难度层级内部选取典型样本词；最后从释义原则与释义用词等维度，对三者所展示的方言词释义进行对比分析。本研究将专用学习词典与大语言模型进行对比，初步揭示了二者在方言词释义处理上的异同。相关发现可为学习者在网络环境下选择方言词查询途径提供参考，也可为融媒词典的后续优化及大语言模型在教育领域的应用探索提供实证依据。

Abstract: Outward-oriented Chinese learning dictionaries are important tools for international Chinese language education, yet the definition of dialectal words in such dictionaries has long lacked a unified standard. The media-integrated dictionary represented by “JUZI Chinese” is attempting to break through traditional compilation models, while Large Language Models, as emerging channels for Chinese language retrieval and learning, have also attracted attention for their language processing capabilities. Both are language resources directly accessible to learners via the internet, but there is still a lack of systematic empirical comparison regarding their differences in defining specific dialectal words. This study takes “JUZI Chinese” as a case of media-integrated dictionary, and DeepSeek and Doubao as two representatives of Large Language Models, conducting a three-way comparison. First, the dialectal words included in “JUZI Chinese” are graded by difficulty. Then, based on the theory of semantic transparency, typical sample words are selected from each difficulty level. Finally, a comparative analysis of the dialect word definitions presented by the three resources is carried out from the dimensions of definition principles and definition wording. By comparing a specialized learner’s dictionary with Large Language Models, this study preliminarily reveals the similarities and differences in their handling of dialect word definitions. The findings may serve as a reference for learners in choosing query channels for dialectal words in an online environment, and also provide empirical evidence for the future optimization of media-integrated dictionaries and the exploration of Large Language Models applications in educational contexts.