DC 欄位 |
值 |
語言 |
DC.contributor | 資訊管理學系 | zh_TW |
DC.creator | 徐子喬 | zh_TW |
DC.creator | Hsu Tzu-Chiao | en_US |
dc.date.accessioned | 2016-8-5T07:39:07Z | |
dc.date.available | 2016-8-5T07:39:07Z | |
dc.date.issued | 2016 | |
dc.identifier.uri | http://ir.lib.ncu.edu.tw:88/thesis/view_etd.asp?URN=103423016 | |
dc.contributor.department | 資訊管理學系 | zh_TW |
DC.description | 國立中央大學 | zh_TW |
DC.description | National Central University | en_US |
dc.description.abstract | 文字探勘技術的進步,使應用於非結構化資料的文字探勘研究在近年迅速增加。隨著文字探勘結合人文社會科學、政治學領域的研究逐漸興盛,專門探討領袖語藝的文獻,卻還停留在以傳統人工閱讀方式進行文字探勘。國內目前應用文字探勘在政治領域的相關研究,受限於作者背景知識,大多無法自行編寫客製化的文字探勘工具。本研究以資料科學的角度與知識背景出發,希望能更為彈性依照研究需求調整文字探勘工具,分析馬英九在中華民國總統任期內的領袖語藝,推論總統語藝、政策走向、重要幕僚任免、處理重大歷史事件態度之間的關聯。
本研究建立一新的研究方法,混合N-gram相連詞出現機率與Word2Vec詞向量的計算,以及用t分布隨機近鄰嵌入 (t-distributed stochastic neighbor embedding, t-SNE) 演算法將200維的Word2Vec詞向量降到2維,並繪製於2維平面等三種技術,輔助研究人員分析總統語藝,尋找隱藏在總統語藝背後的論述、思想及決策脈絡。
研究結果指出,將每篇文本分開做文字探勘,觀察文字探勘數據分析結果的變異並紀錄變異的時間點,再從發生變異的時間點著手研究變異原因,可以找出馬英九總統語藝和政策走向、重要幕僚任免、面對重大歷史事件態度間的關聯。根據分析結果並調查相關證據,政策走向、重要幕僚任免部分有兩大發現,一是2013年10月開始馬英九政府的區域整合政策由兩岸走向全球,由劉大年擔任政策幕後推手,劉大年也在2014年5月走馬上任擔綱國安會副秘書長一職,從幕後走向幕前。二是從2013年10月習近平提出兩岸領導人會面構想開始,馬英九在接下來的三次國慶演說講稿中詞向量最接近「兩岸」的詞之中都有「和平」一詞存在,直到兩岸領導人會面當天馬英九的開場談話也以兩岸和平為主軸,顯示馬英九想要留下「兩岸和平」的歷史定位。面對重大歷史事件態度部分,分析結果則指出,馬英九政府在最後兩年任期,對二二八事件的態度轉為應卯。
最後探討資料集一致性問題影響文字探勘數據訓練的可能性,從研究過程推斷總統語藝的文字探勘比較適用分篇文字探勘的研究方式。建議未來的研究者在探勘總統語藝時,採用單篇文本分析,而非全部文本綜合在一起分析的策略。 | zh_TW |
dc.description.abstract | Recent trends have been influenced by the progression of text mining so researches on unstructured data mining increase rapidly. There are many literatures on using text-mining techniques for politics but studies of rhetoric of leaders are still analyzed textual materials via reading by human. This paper attempts to apply text-mining techniques to research of rhetoric of the former president Ma Ying-jeou and find the relationships between rhetoric and policy, staffing assignments or attitude toward important historical events.
In this paper, a new research method to analyze and reveal hidden ideology and decision-making process behind rhetoric of the president is developed. We present here an original work that employs Word2Vec, N-gram and t-distributed stochastic neighbor embedding (t-SNE) to assist rhetoric analysis.
The results show that mining each document sorted by date on rhetoric of the president, observing variations in data calculated by text mining techniques, and then investigating the reasons why data alter on the time can find out the association between rhetoric and policy, staffing assignments or attitude toward important historical events.
Furthermore, we disclosure the mastermind behind the scenes of regional economic integration policy was Daniel Liu(劉大年)in the period from October 2013 to May 2016, Ma Ying-jeou tried hard to make his historical role as a man bringing Cross-strait peace and the attitude of the Ma administration was perfunctory toward February 28 Incident after February 2015. | en_US |
DC.subject | 領袖語藝 | zh_TW |
DC.subject | 文字探勘 | zh_TW |
DC.subject | 馬英九 | zh_TW |
DC.subject | Word2Vec | zh_TW |
DC.subject | N-gram | zh_TW |
DC.subject | t-SNE | zh_TW |
DC.subject | Word2Vec | en_US |
DC.subject | N-gram | en_US |
DC.subject | t-SNE | en_US |
DC.subject | rhetoric of the leader | en_US |
DC.subject | text mining | en_US |
DC.subject | Ma Ying-jeou | en_US |
DC.title | 總統語藝之文字探勘研究 | zh_TW |
dc.language.iso | zh-TW | zh-TW |
DC.title | Text Mining on Rhetoric of the President | en_US |
DC.type | 博碩士論文 | zh_TW |
DC.type | thesis | en_US |
DC.publisher | National Central University | en_US |