Please use this identifier to cite or link to this item: https://ah.nccu.edu.tw/handle/140.119/118960


Title: 以大數據分析影響唐詩流通度之因素
Using big data to analyze the reasons for the popularity of Tang poetry
Authors: 黃泰霖
Huang, Tai-Lin
Contributors: 宋傳欽
姜志銘

Song, Chwan-Chin
Jiang, Jyh-Ming

黃泰霖
Huang, Tai-Lin
Keywords: 大數據
唐詩
流通性
主成分分析
因子分析
詞嵌入法
Big data
Tang poems
Popularity
Principal component analysis
Factor analysis
Word embedding method
Date: 2018
Issue Date: 2018-07-27 12:13:54 (UTC+8)
Abstract: 本研究旨在探討唐詩在流通上的特性與原因,期望能為唐詩詩學研究提供新的研究方向。本文以《唐詩排行榜》所建立的資料作為出發點,並以主成分分析與因子分析為主要的分析方法,萃取出唐詩在流傳上的特性及因素,探討古人與今人在詩文閱覽偏好的不同,並進一步利用詞嵌入法探討詩文內容相似度與主成分分析及因子分析之結果在排序上是否一致。
經過對唐詩排行榜數據的研究,本文發覺主成分分析總結出以下兩項特性:1. 時代性差異 2. 詩文收錄完整性,其中時代性差異顯示『每一個時代的前理解不同,審美標準自然有明顯落差,因而造成古今閱眾對於詩文的欣賞與偏好有一定程度的差異』;而詩文收錄完整性指的是『隨著編纂需求的不同,詩作在流傳上可分為 1. 完整詩文 2. 片段名句 兩種類型』。
而因子分析則總結出兩個影響唐詩流通的原因:1. 歷史性強度 2. 詩學經典性,其中歷史性強度所代表的是『古今閱眾在詩文內容的喜好上,深受詩文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,可區分詩文是否為一派之經典』
利用詞嵌入法進行詩文文本的相似性研究,發現第一主成分時代性差異、第一因子詩學經典性以及第二因子歷史性強度之結果與其分別對應之詩文相似度排序具有顯著的一致性。
This study aims to explore the characteristics of the popularity of Tang poetry, and hopes to provide new research direction for Tang poetry. First, we use multivariate statistical methods, which include principal component analysis and factor analysis, to analyze the data given by the book Ranking on Tang Poems. Based on the results of analysis, we extract the characteristics of the popularity of Tang poetry, and compare modern with ancient preferences of reading. Finally, we use word embedding techniques to further analyze the suitability of the results extracted by principal component analysis and factor analysis.

After analyzing the data given by the Ranking on Tang Poems, principal component analysis suggests the following two characteristics: time difference and poem integrity. “Time difference” refers to “Having its own pre-understanding, each era has its own aesthetic standard, which makes some differences of poetic appreciation between ancient and modern readers”. “Poem integrity” refers to “A poem is selected either in a complete form or in a partial form according to the editing requirements.”

Based on factor analysis, we sum up two factors that may influence the popularity of Tang poetry: history related strength and poetic classicism. The “history related strength” refers to “The poem preferences of ancient and modern readers may be influenced by the history related strength of the poem.” The “poetic classicism” indicates that “Poem can be considered to lead a school of thoughts from the academic perspective.”

Using word embedding techniques to study the textual similarity of poems, we find that each of first principal component and two factors has a significant rank correlation with the textual similarity of the top ranking poems based on its corresponding principal component or factor.
Reference: Gao,J.(2018).Chinese-poetry. https://github.com/chinese-poetry/chinese-poetry.
Johnson, R. and Wichern, D.(2007). Applied multivariate statistical analysis(6th ed.). Prentice Hall, Upper Saddle River, NJ.
Le, Q. V. and Mikolov, T. (2014). Distributed representations of sentences and documents. Computing Research Repository, arXiv:1405.4053.
Mikolov, T., Chen, K., Corrado, G., and Dean, J. (2013). Efficient estimation of word representations in vector space. Computing Research Repository, arXiv:1301.3781.
Řehůřek, R. and Sojka, P.(2010). Software Framework for Topic Modelling with Large Corpora. In Proceedings of the LREC 2010 Workshop on New Challenges for NLP Frameworks, pages 45–50, Valletta, Malta. ELRA. http://is.muni.cz/publication/884893/en.
王兆鵬、張靜、邵大為、唐元 (2011)。唐詩排行榜(初版)。北京:中華書局。
王宏林 (2012)。論唐詩經典的基本屬性,建構要素及途徑。許昌學院學報,31(4):54,58。
蔣寅 (2003)。中國古代文學通論隋唐五代卷(初版)。遼寧:人民出版社。
趙義山、李修生 (2010)。中國分體文學史詩歌卷修本(2版)。上海:上海古籍出版社。
陳耀茂 (1999)。多變量解析方法與應用(初版)。台北:五南圖書出版公司。
魯迅 (2005)。魯迅全集第 13卷(初版)。北京:人民文學出版社。
Description: 碩士
國立政治大學
應用數學系
104751013
Source URI: http://thesis.lib.nccu.edu.tw/record/#G0104751013
Data Type: thesis
Appears in Collections:[應用數學系] 學位論文

Files in This Item:

File SizeFormat
101301.pdf1524KbAdobe PDF0View/Open


All items in 學術集成 are protected by copyright, with all rights reserved.


社群 sharing