Applications of Text Mining to Studying the Association between Responses and Opinions from Opinion Web System of the National Taiwan University

隨著網路資訊發達及行動通訊的重度使用發展趨勢，大眾們在網路上留下大量的數據，代表著大數據時代（Big Data）的來臨。根據國際數據資訊中心IDC公司統計，2020年時全球總資料量將到達40 Zettabyte（ZB），相當於約43兆Gigabyte（GB），相較於2010年時超過五十倍的成長，且文字、圖片、視頻及音頻等非結構化資料的應用也會越來越頻繁。其中網路勢力崛起的鍵盤力量，讓文字儼然成為網路世界中大家溝通討論的媒介，重要性不可或缺。 因此，在現今社會中除了利用量化的資料進行分析外，質化的資料含有更大量的資訊，其分析的結果也更具備價值性。故本研究延續「國立台灣大學校務會議及校務建言資料之分析研究」量化性的研究結果，進一步對台大校務建言的內容進行質化的資料分析。 透過文字探勘（Text Mining）進行中文斷詞、潛在語意分析（Latent Semantic Analysis）及情緒分析（Sentiment Analysis），從眾多繁雜、尚未處理的文字中，找出被隱藏在字裡行間的重要資訊。來探討學生使用校務建言系統來表達意見，究竟為了什麼樣的溝通目的及需求；校方面對這樣的問題及建言，究竟如何回應學生，其是否有真正回答並處理問題，還是僅僅只是敷衍學生罷了。再者，兩方在溝通的過程中是否確實地落實真實的雙向「理性」溝通。 在校務建言的橋樑中，若能透過文字探勘的分析，從中探討雙方在溝通上的問題並給予建議，使學生更能以理性的思辨與態度，提出建言及問題；學校更能以積極的誠意與態度，處理及回應建言。讓行政單位和學生透過校務建言這個網路意見交流的平台，彼此間有良性的互動溝通，這將是使學校的運作有更好的發展。Advanced network information and growing mobile communications have resulted in an increase of publicly available data on the internet. This indicates the arrival of Big Data. According to statistics from the International Data Corporation （IDC）, the overall volume of data worldwide will reach 40 Zettabytes（ZB）or, 43 trillion Gigabytes（GB）by 2020. This will generate a 50-fold growth from 2010. The frequent use of unstructured information such as text, images, video and audio will also become greater. Specifically, the rise of the power of keyboard has made text-based communication an essential channel for the public to discuss and exchange information online. Aside from the commonly used quantitative analysis, qualitative data incorporates extensive information to provide additional value of analyzes. This study follows Yao’s work: “Statistical Analysis of the Data from University Assembly Meetings and the Opinion Web System of the National Taiwan University” and utilizes its quantitative analysis results to further conduct qualitative analysis on the National Taiwan University’s opinion web system. This research aims to search for hidden information in complicated unprocessed text through text mining which involves text segmentation, latent semantic analysis and sentiment analysis. By using these approaches, it examines the issues that students used the system to express their opinions; and whether the responses from the university effectively and adequately responded and resolved these issues. The research also examines whether both sides have actually communicated in a rational manner. To better the communication between students and the university on the opinion web system; this research used the technic of text mining to uncover the problems that occurred in the process of information exchange. It gives further recommendations for students to raise questions through rational and critical thinking and for the university to respond with a positive and genuine attitude. This can enhance the operations of the university and lead to a better development in the future