根據尼爾森2013年的市調發現，約77%的消費者相信陌生網友在網路上發表的意見評價。從語音辨識技術發展趨勢來看，越來越多應用於家電、智慧型手機，甚至在公共建設上等。例如Apple推出的「Siri」，藉由語音輸入進行人機互動，自動地為使用者處理事情，像是排定會議、撥打電話及傳送訊息，甚至能與使用者對答，至今仍令人津津樂道。透過語音辨識輸入，可增加蒐集及尋找資料的效率和速度。 目前語音辨識多應用於語言學習，如英文、日文之發音練習，也有一些應用於物聯網領域，然而卻鮮少看見與電子商務及輿情分析結合的相關例子。據此，本研究嘗試結合語音辨識及輿情分析技術，建置一個雛型推薦系統，自動化擷取網路上的討論文章以及產品資訊，進行關鍵字擷取、情感分析等步驟，整合成基於語音辨識之電子商務輿情推薦雛形系統，透過使用者語音的輸入，來推薦較符合使用者需求之產品。 本研究讓使用者體驗雛型系統後進行問卷調查，以獲取使用者的意見反饋，並篩選出可行的建議，作為未來系統發展之參考，目的是找出較符合使用者期待之語音辨識電子商務模式。本研究發現68%的受測者是較喜歡輿情導向模式，對系統平均滿意度為4.2分(滿分為5分)，期望本研究結果能對未來相關領域有所貢獻。 According to market investigation report by AC Nielsen in 2013, about 77% of consumers are swayed by opinions and comments posted on the internet by other users.On the perspective of speech recognition development, more and more examples are applied to home appliances, smart devices and public utilities. For example, a system developed by Apple, called “Siri”. It can interact with users through speech recognition inputs, manage user schedules, conference arrangement, phone calling and message sending, or even chat with the user in ways. Now is still widely used. By speech recognition input, it can increase the efficiency and speed of information searching. Speech recognition techniques are mostly applied to language learning like English and Japanese pronunciation practice. Some are applied to internet of things. However, we rarely see the examples of combination with E-commerce and sentimental analysis. Thus, this study attempts to combine the speech recognition and sentimental analysis technique to build a prototype of E-commerce recommendation system. This prototype system automatically collects articles and information of products on the web, catches keywords and sentimental analyzing, and identifies the input of user with speech. Then recommends to user for the best product as a result. Users were asked to take a survey after experiencing a prototype system built for this study. We took feasible feedback as future reference for developing the system. The purpose of this study was to find the most expected speech recognition e-commerce mode of users. The research found that 68% of users prefer the sentiment-oriented mode to others. The average satisfaction level of system is 4.2 (out of 5). With this data, we hope to contribute more to the development and implementation in this field of study.