創新的社群文字探勘方法分析2018台北市市長候選人形象定位

大數據的時代來臨，大量的資料充斥於我們生活當中，若能有效的分析，即能瞭解消費者、市場以及趨勢，創造極大的價值。近年來社群媒體興起，於社群媒體上的數據分析受到矚目，人們習慣於社群平台上分享主見、看法及偏好，於政治領域的影響更為顯著，網民的言論與候選人的形象息息相關。因此，若能有效分析這些非結構化的資料，即能成為評估候選人的重要參考指標。大數據與文字情緒分析方法愈趨成熟，過往有學者運用文字情緒方法評估政治候選人，主要是以英語的應用為基礎。本研究針對2018台北市市長選舉三位可能候選人為例，以文字探勘與情緒分析方法，先建構中文政治候選人形象特徵之關鍵字詞庫，針對三位候選人粉絲專頁上選民的評論與意見，運用情緒分析評估候選人於五個形象構面上的表現加以分析。此創新的文字探勘與情緒分析調查方法，提供傳統民調另一種聲音與角度的參考，並提供競選團隊參選策略的參考依據。

關鍵字

文字探勘；情緒分析；候選人形象

並列摘要

As to the era of big data, the volume of data has exploded in our lives. If data can be analyzed effectively, the results will create significant values in business. Recently, as social media blooming, people are used to sharing opinions on social media. Therefore, the data analytics for social media is in the spotlight. Furthermore, the effect of social media is significant in the field of politics because internet comments are closely related to the image of politicians. If unstructured data can be analyzed, the results will become important in evaluating politicians. Big data and sentiment analysis are developing, researchers have applied sentiment analysis to evaluate politicians based on the English Lexicon but none of them used Chinese Lexicon to conduct sentiment analysis. This study focuses on the candidate analysis of Taipei City Mayor Election in 2018. For the research, the researcher establishes the Chinese Characteristics Keyword Lexicon on the basis of sentiment analysis, analyzes the internet comments on the Facebook fan pages of three candidates and calculates text sentiment scores on the five dimensions in order to evaluate the performance of politicians on each dimension. The innovative way of data mining and sentiment analysis for politicians' evaluation provide different voices and perspectives for campaign teams to map strategies.