由於網際網路的快速發展，許多服務與購物網站累積了許多評論資訊。這些線上網站的評論及評價方式皆是依據各網站自我事先的定義，無法針對每位使用者個別的特殊需求，讓使用者找到自己所關心的相關評論資訊。因此本研究提出新的評論分析方法，讓使用者針對自己考量的訂房因素，並讓使用者提供相關的關鍵字詞，進一步進行客製化的評論分析。透過本研究我們可以依據使用者的需求產生評論分析，將評論內容分為不同構面以及情緒傾向，藉此得知整間旅館在多篇評論中所表現意見。本研究以hotels.com上的旅館評論作為分析之資料，實驗分為以下兩個部份，第一部份評估構面及情緒分析方法之效果，第二部份實際將不同旅館的評論分群以獲得評論中的意見。 As the rapid development of Internet, many services and shopping sites accumulated a lot of comment information. However, these websites defined specific formats of their own to categorize or to show all of the users’ reviews which users cannot easily find out information they really care about in tons of reviews on the website. Therefore, this study proposes a new review analysis method, allowing users to consider for their own reservations factors and to provide relevant keywords, to customized review analysis for specific user. According to the users’ need, we classified the content of reviews into different dimensions and different sentiment tendencies in review analysis. Through this research can be generated based on the user's needs in review analysis, and further be able to tell the whole hotel’s opinion which is presented in many reviews. Reviews on hotel.com are used in this paper. The experiments in this paper are as follows. First, evaluate the effect of dimension analysis and sentiment analysis. The second part is to put reviews on the website in clusters and to receive the specific opinions within all the reviews.
