侯文娟Hou, Wen-Juan楊登堯Yang, Deng-Yao2019-09-052017-09-012019-09-052017http://etds.lib.ntnu.edu.tw/cgi-bin/gs32/gsweb.cgi?o=dstdcdr&s=id=%22G060447053S%22.&%22.id.&http://rportal.lib.ntnu.edu.tw:80/handle/20.500.12235/106467過去人們獲取資訊的途徑只有從談話、書籍、報章雜誌等媒體，資訊量的收集速度緩慢且數量有限，然而現今網路的發達以及科技改良所賜，網路的方便性及發達帶給了這個社會資訊化。 社群網站的興起(例如:facebook、twitter)，讓許多人開始透過這些網路平台，迅速傳播新聞資訊或就生活上的知識進行交流與溝通。報紙雜誌等傳統媒體，也開始透過網路平台進行發佈。 然而在資訊爆炸的時代，人們該如何從這些大量的報導中獲取想要或者喜歡的資訊，而媒體又該如何從新聞內文中適當的撰寫以便吸引閱聽人，讓人們能夠喜歡閱讀該報導，並且可以從這些報導當中，發掘現今人們的新聞喜好傾向，這些都是目前值得研究者探討且著墨的地方。 本研究將首先利用情緒分析的技術，分析現在網路新聞內文經常使用哪些詞彙或語句，可激發閱聽人的情緒反應以增加其閱讀興趣。其次，為了更進一步了解新聞極性的趨勢，也就是正向的新聞比較受歡迎還是負向的新聞比較受歡迎，會先進行斷詞之後，利用TF-IDF值尋找出關鍵字，然後利用語料庫進行比對，得到正向詞與負向詞的資訊，接著再利用Facebook提供的讚數當作佐證，就可以看出現在人們是喜歡哪一類的新聞。 研究結果發現，閱聽者比較常關注負向新聞，並且本研究利用圖斯勒(Marc Trussler)和索羅卡(Stuart Soroka)在加拿大麥基爾大學(McGill University)的實驗結果相比對，顯示和從心理系角度所做的研究，有相符的結果，進而佐證本研究的可信度。In the past, the way of people obtaining information is only from the conversation, books, newspapers and other media information collection, which is slow and limited in number. However, due to the development of network technologies, in the present, the vast amount of information can be retrieved conveniently from internet. Some community web sites (such as facebook , twitter)make many people start with these network platforms for the rapid dissemination of news or for exchange of knowledge on life. Traditional newspapers, magazines and other traditional media also begin to publish their reports on the network platform. In an era of information explosion, how people can get information that they want or like from these extensive reports and how news can attract readers are worthy of investigation. Furthermore, the preference tendency of negative news or positive news is also very important. The study will first take advantage of the sentiment analysis technology to analyze network news by extracting the frequently-used words or phrases in order to increase people’s interest in reading. Second, to further understand the trend of news polarity, that is, whether positive news is more popular than negative news or not. The study segments words, finds keywords using TF-IDF values, and then matches keywords with a sematic dictionary in order to get the polarity informations. Finally, use the number of ’like’ provided by Facebook as corroboration, the trend of news polarity that people like is shown. The study shows that readers more concerned about the negative news. Comparing to the psychological research of Trussler and Soroka in Macgill University in Canada, the result is consistent. It thereby furthermore support the confidence of this study.自然語言處理情感分析中文剖析器語意字典NLPsentiment analysisChinese parsersemantic dictionary