利用LSTM建立聲音滿意度辨識模型;Construct a vocal satisfaction identification model with LSTM

NCU Institutional Repository > 管理學院 > 企業管理研究所 > 博碩士論文 > Item 987654321/83363

jsp.display-item.identifier=請使用永久網址來引用或連結此文件: http://ir.lib.ncu.edu.tw/handle/987654321/83363

题名:	利用LSTM建立聲音滿意度辨識模型;Construct a vocal satisfaction identification model with LSTM
作者:	沈依;Shen, Yi
贡献者:	企業管理學系
关键词:	電話行銷;顧客滿意度;LSTM;Telephone marketing;Customer Satisfaction;LSTM
日期:	2020-07-15
上传时间:	2020-09-02 15:31:17 (UTC+8)
出版者:	國立中央大學
摘要:	隨著行動電話的蓬勃發展，電話已儼然成為行銷的強大工具，並且以電戶行銷的方式可客製化的制定適合的策略以達到顧客滿意，藉此特性每日皆會有大量地聲音資料產生。然而，對於電話行銷而言，若要探討顧客滿意度，無法使用傳統紙本問卷之形式蒐集，必須透過語音對話的方式進行探討，而目前已有學者透過語音資料以情感或是情緒探討滿意度，但無論是情緒或情感皆是多元且複雜的，其正負向之情感或情緒皆不能代表顧客的滿意程度，因為正向可能包含了開心、快樂、滿意，負向可能包含沮喪、難過、不滿意，而滿意以及不滿意皆只是其中可能的結果，且目前尚未有學者利用真實的客服中心語音資料建立滿意度辨識模型。故本研究將建立一套LSTM聲音滿意度辨識模型，結合真實客服滿意度調查語音資料與顧客購買之歷史紀錄，以探究顧客真實的滿意度感受。而本研究首先以實驗設計蒐集實驗室資料以建立實驗室滿意度辨識模型，並以此模型當作業界模型之雛形，再以相同的方式建立業界滿意度辨識模型。然而，最後實驗結果得出實驗室與業界滿意度辨識模型精確率皆高於70%，依此結果，此模型可有效地了解顧客之滿意程度，並可掌控電話行銷進行之時間，以降低成本及提升客服人員服務品質管理以及顧客關係管理的維護，使企業增加效能與利益。;As the rapid development of technology, smart phone becomes a powerful tool for marketing and telephone marketing can customize the suitable strategies to make customer feel satisfied, thus it would generate a large amount of vocal data every day. However, if we want to know the satisfaction of telephone marketing, we must use the vocal data instead of paper questionnaire. There are some scholars using vocal data to discuss satisfaction by emotion or sentiment and divide it in to positive and negative, but whether the emotion or sentiment are diverse and complex, the positive may include happy, excited and satisfied and the negative may include frustrated, anger and dissatisfied, according to the above, satisfied and dissatisfied are just the possible results, so we can’t use it to represent the satisfaction of the customer, and no scholars have used the real call center’s vocal data to construct a satisfaction identification model. Therefore, this study will construct a vocal satisfaction identification model with LSTM, it combines with real vocal data of satisfaction survey and the history records of customer purchase, to explore the real customer satisfaction. In this study, the identification model derived from the vocal collected in the lab serves as a prototype to develop the satisfaction identification model for vocal collected from call centers. Based on the result, the accuracy of these two models are higher than 70%. To sum up, this model can effectively identify customer satisfaction and control the duration of a call, it can reduce costs, improve the service quality, maintain customer relationship, enhance company’s efficiency and benefits.
显示于类别:	[企業管理研究所] 博碩士論文

文件中的档案:

档案	描述	大小	格式	浏览次数
index.html		0Kb	HTML	92	检视/开启

在NCUIR中所有的数据项都受到原著作权保护.

社群 sharing

数据加载中.....