Please use this identifier to cite or link to this item:
標題: 一種非干擾式顧客滿意度偵測方法與其應用之研究
A Non-intrusively Detective Method with Empirical Study for the Evaluation of Customer Satisfaction
作者: Jang, Yu-Teng
關鍵字: 顧客關係管理;Customer relationship management (CRM);顧客滿意度;服務補救;訊號處理;語音辨識;功率頻譜密度;類神經網路;Customer satisfaction;Service recovery;Signal processing;Voice recognition;Power spectrum density (PSD);Artificial neural network
出版社: 電子商務研究所
顧客滿意度在企業體的顧客關係管理(CRM)議題中,一直都扮演了一個很重要的角色。 然而現行顧客滿意度衡量方法,大多採用訪談或發放研究問卷的方式來進行,而其所回收到的結果,也越來越少(原因是現在的顧客,很少人願意花時間填寫問卷,而且就算填寫了問卷,也不容易衡量問卷結果的準確性,當然也很難判斷顧客是否說謊或是草率回答,因此容易產生衡量偏差)。 在此研究中,我們提出了一個創新性非干擾式的滿意度衡量方法,也就是直接以顧客說話的聲音,進行該顧客之滿意度衡量,並建議能作為補足傳統滿意度衡量方法缺點之新方法。由於受到了人工智慧(AI)相關領域中之機械學習程序研究的啟發,我們將具代表性的聲音資料經過轉換處理,並採用合適的倒傳遞式類神經網路(BPN)加以訓練,此智慧型系統將能夠只藉由顧客說話的聲音,就直接衡量出該顧客內隱的滿意度。經過了三個主要的實驗與實地測試,我們的實驗結果顯示,此新方法的準確度無關於語句、性別、或受測者,且平均可達到約80%的準確度,並在實驗室以外的實際測試中,發現此方法的準確度(平均約83%)甚至比人類判斷的準確度(平均約62.5%)還要來的高。在我們的系統架構中,我們進一步建議可以根據滿意度衡量結果而給予該顧客即時適當的回饋。由於此方法,是直接以顧客聲音來進行衡量與判斷(不需面對面),因此將可大幅利用於電話訪問、語音化網站系統、語音電子商務平台、手機電信服務等範疇。適當地應用這個方法,將能夠達到提升顧客滿意度、提高顧客忠誠度、增加顧客再次購買率,以及提升企業利益及獲利等目標。且此方法除了可以延伸到其他領域範疇並提升該領域的商業價值外,其實驗結果也為相關領域的研究與發展播下種子,並成為未來研究發展所需之重要參考資料。

In this research, we propose a non-intrusive method that can be integrated in a voice-enabled web system for evaluating customer satisfaction directly via his/her voice or work alone. Based on this approach with experimental and empirical results, we recommend that this method may be treated as a new way to complement the traditional methods which evaluate satisfaction through surveys of using questionnaires. We carried out three major experiments and conducted an empirical study based on the neural network trained by using a collected voice record database, and obtained semantic-independent, gender-independent, and speaker-independent results showing an average accuracy rate of approximately 80% in each experiment. In our empirical study, we find that computers (with the average accuracy 83%) perform better than human (with the average accuracy 62.5%) on satisfaction detection directly by voice; especially the voice record is only a piece of sentence. The recognition rate reveals the effectiveness and feasibility of the proposed method. In addition, a system framework with the mechanism of giving immediate proper feedbacks (such as an unexpected gift) to customers is discussed and can be applied on the proposed voice-enabled web system. With this powerful technique, direct managerial implications (such as improving customer satisfaction, enhancing customer loyalty, increasing re-purchase rate, and improving enterprise's benefits and profit) are no longer impractical. Feasible marketing application and further research perspectives are presented at the end.
Appears in Collections:科技管理研究所

Show full item record

Google ScholarTM


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.