習近平時期中國對台政策話語的演變（2012-2024）：以文字探勘為研究途徑

Abstract

本研究探討習近平時期（2012-2024）中國共產黨對台政策話語的語言特徵與結構性變化。隨著兩岸互動模式與外部政治環境的轉變，中共涉台論述在語言配置與敘事重心上出現顯著調整。為降低主觀詮釋與分期偏誤，本文結合文字探勘方法與批判話語分析（Critical Discourse Analysis, CDA）之三層次架構（語意層、語境層、結構層），系統性分析《人民日報》於 2012 至 2024 年間共 4,454 篇涉台報導。 研究方法採取由下而上的資料導向設計，包含斷詞與 TF-IDF 權重計算以辨識關鍵詞彙、LDA模型(Latent Dirichlet Allocation)進行主題建模，以及詞彙共現與社會網絡分析以建構概念間的語意與語境關係。研究結果顯示，在語意層面，「和平」相關語彙持續出現，但其使用逐步附帶明確政治前提，「台獨」與「外部勢力」相關詞彙之權重與指認頻率則明顯上升。在語境層面，台灣主要政黨與民主相關概念呈現差異化配置，顯示其語言功能隨時期而調整。在結構層面，對台話語於習近平第一任期仍保有一定彈性，第二任期以後則更集中於民族復興與國家安全的整體敘事。 本文指出，《人民日報》的涉台報導不僅反映政策立場，亦透過制度化的語言運作參與政治問題的界定。本文以文字探勘輔助話語分析，提供理解習近平時期中共對台話語變遷的實證基礎。

This study examines the linguistic characteristics and structural transformation of the Chinese Communist Party’s (CCP) Taiwan-related policy discourse during the Xi Jinping era (2012-2024). As cross-strait interactions and the external political environment have evolved, notable adjustments can be observed in the configuration and narrative focus of Taiwan-related discourse. To reduce subjective interpretation and periodization bias, this study integrates text mining techniques with the three-layer framework of Critical Discourse Analysis (CDA) the semantic layer, contextual layer, and structural layer to systematically analyze 4,454 Taiwan-related articles published in the People’s Daily between 2012 and 2024. Methodologically, the study adopts a bottom-up, data-driven approach. This includes word segmentation and TF-IDF weighting to identify salient keywords, topic modeling using Latent Dirichlet Allocation (LDA), and word co-occurrence analysis combined with social network analysis to construct semantic and contextual relationships among key concepts. The findings indicate that, at the semantic level, references to “peace” persist over time but are increasingly accompanied by explicit political preconditions, while terms related to “Taiwan independence” and “external forces” show a clear rise in both weight and frequency of identification. At the contextual level, Taiwan’s major political parties and democracy-related concepts exhibit differentiated configurations, suggesting that their discursive functions vary across periods. At the structural level, Taiwan-related discourse during Xi Jinping’s first term retains a degree of flexibility, whereas after the second term it becomes more consolidated within overarching narratives of national rejuvenation and national security. This study argues that Taiwan-related reporting in the People’s Daily not only reflects policy positions but also participates in the discursive delineation of political issues through institutionalized language practices. By employing text mining as an analytical aid to discourse analysis, this research provides empirical evidence for understanding the transformation of CCP Taiwan-related discourse during the Xi Jinping era.