元和體와 西崑體에 관한 小考 ― 텍스트 마이닝 분석 기법의 접근

Abstract

This paper examines Tang Dynasty Yuan-he Style and Northern Song Dynasty Xi-kun Style from a lexical perspective. As the amount of texts analyzed for each style is huge and previous studies require a different approach to examine the texts analyzed by qualitative analysis, this study attempted a digital humanities approach. I attempted a quantitative approach through analysis based on n-gram analysis of representative text mining techniques, cosine similarity analysis using its results, TF-IDF figures, and topic modeling through LDA models. Based on the comparison of the frequency of single-letter unigrams and two-letter bigrams and their TF-IDF values, topic modeling, and other criteria, it was found that Yuan-he style frequently and heavily used casual vocabulary and people-related vocabulary, such as Ren(人) and Xin(心), which is related to the realistic features of Yuan-he style, which is socially participatory and people-oriented. The frequency of words related to splendour, such as yu(玉) and jim(金), was high in Xi-kun style, and words that do not appear in the top 500 bigram frequencies of ‘The Complete Song Poems’, were also high in Xi-kun style, such as Songyu(宋玉), Xiangru(相如), Handian(漢殿=漢宮), and Liangyuan(梁園=梁苑), were the most common. This shows that a poet's poetic background has a significant influence on the choice of vocabulary. On the other hand, based on the cosine similarity analysis, we found that the major poets of a school did not necessarily use similar vocabulary. Yuan-zhen and Bai-Juyi, two of the key poets of the Yuan-he style, were more than 30% similar in their vocabulary use. In contrast, no lexical connections were found between the three major Poets of the Xi-kun style, Yang-yi, Liu-yun and Qian-weiyan, which is a surprising conclusion. The statistics used in this article are obviously subject to error. However, we can argue that they are valuable in that they allow us to effectively analyze things that are not easily accessible to humans by hand.