隨著 Web2.0 的興起,網路使用者已習慣將自己的消費經驗記錄在網路上,這些分享即是所謂的網路口碑。而因為參考網路口碑可以降低消費者購後失調的風險,所以是消費者於購買前常會參考的重要來源。然而目前的網路口碑推薦系統,因為對網路口碑無法擁有資訊的主導權,所以都只將網路口碑的效益發揮於多數人已經消費之後,造成系統只能推薦舊的商品給消費者,可是消費者往往需要被推薦的卻是新的商品。因此本研究提出利用本體技術管理網路口碑之大眾經驗,以相同類型商品之商品屬性,建立消費事件間的關聯,使過去的網路口碑可以用來預測新商品是否值得購買。而最後的驗證是以本研究提出的預測結果與消費者最終接受的結果進行比較,經由驗證結果顯示本研究提出的預測方法之準確率高達 73.33% ,雖然比起人為過濾之準確率 80% 略低一點,但其應該能增加決策的效率,並減少目前消費者協同預測方式於人力上的浪費。
For users of recommendation system, obtaining recommendation list with new products are more useful than old products. The recommendation based on electronic word-of-mouth (eWOM) usually offers old products to consumer, because the recommendation has no enough eWOM in initial. This study is devoted to use eWOM of old products to predict rating of new products to consumer. Different from traditional approaches that treat reviews of one consuming event as a whole; our approach considers the relation between these consuming events. The mechanism is using eWOM to evaluate each attribute of old product, and takes ontological techniques to construct the relation between old products and new products of the same kind by the attributes. By matching final good rating of new product, experiment showed that the accuracy of our system was 73.33%. The result was lower than 80% accuracy of the collaborative predicting, but we could conclude that the proposed scheme compared with collaborative predicting should help decrease the cost of human resource and time.