唐詩流通度之探討

本研究旨在探討唐詩在流通上的原因，期望能爲唐詩詩學提供新的研究方向。本文以《唐詩排行榜》一書所建立的100首詩作之資料作爲出發點，並以因子分析爲主要的分析方法，萃取出唐詩在流傳上的因素，探討古人與今人在詩作閱覽偏好上的差異，並進一步利用詞嵌入法探討詩作在因子分析結果上之排序是否與在詩作間相似度上之排序一致。經過對唐詩排行榜數據的研究，本文根據因子分析之結果，總結出兩個影響唐詩流通的原因：歷史性強度與詩學經典性，其中歷史性強度所代表的是「古今閱衆在詩作內容的喜好上，深受詩作內容的歷史背景所影響」；而詩學經典性則顯示「從詩學學術領域的角度出發，可區分詩作是否爲一派之經典」。利用詞嵌入法進行詩作的相似性研究，我們發現詩作依第一因子「歷史性強度」以及第二因子「詩學經典性」之排序結果，都與依據詩作間相似度之排序結果有顯著甚至高度顯著的一致性。

關鍵字

唐詩；因子分析；詞嵌入法；流通性

並列摘要

This study aims to explore the reasons for the popularity of a specific collection of Tang poems and introduce a new quantitative research direction for Tang poetry. Our approach includes two procedures. We first use factor analysis to analyze the data provided in the book ＂Ranking of Tang Poems＂, which is followed by interpreting the extracted factors based on literary theory regarding the popularity of Tang poetry. Then we use word embedding techniques to further justify the suitability of the results extracted by factor analysis. After deciding a two factor solution to the factor analysis, we interpret the two factors that may explain the popularity of Tang poetry to be ＂history related strength＂ and ＂poetic classicism＂. A poem with high ＂history related strength＂ makes references to well-known historical events. ＂Poetic classicism＂ indicates that the poem can be considered as a classic to a certain literary school of thoughts from the academic perspective. Using word embedding techniques to study the textual similarity among poems, we find that each factor has a significant or highly significant rank correlation with the textual similarity, which is based on the top five ranking poems of the corresponding factor.