VoiceXML中文語音系統在電子商務網路環境之技術研究

Abstract

VoiceXML（Voice eXtensible Markup Language）是由AT&T、IBM、Lucent與Motorola所共同制定發起的語音標記語言規格，並通過W3C審核與正式推展。VoiceXML為代表人機介面對話的一種標注語言，而不同於傳統一般HTML語言的是，VoiceXML以語音瀏覽器(Voice Browser)與聲音的輸出及輸入來呈現。 本論文將運用VoiceXML的規格與技術，實做出一個完整的中文語音對話系統，包含自動語音辨識（ASR，Automatic Speech Recognition）、文字轉語音（TTS，TTS, Text-To-Speech）技術以及整合目前日趨成熟的VoIP（Voice over IP）網路電話技術，使整個系統能完全與PSTN整合。另外，本論文運用所提出的以XML為架構之VoiceXML Browser，可讓使用者透過傳統電話、手機或是網際網路，來與本系統做語音查詢與瀏覽網頁之功能。進而突破目前VoiceXML只運用在純語音查詢資料或是語音訂單的應用。 因此，本論文提出一種整合電子商務與VoiceXML的架構，除了結合VUI（Voice User Interface）與GUI（Graphical User Interface）的人機互動式介面特色，更簡化傳統IVR的複雜語音程序以及降低購買語音設備之成本，將有助於降低客戶服務支援中心之成本，並同時提升客戶滿意度。對於使用者而言，本系統也提供更多存取管道，並藉由互動式的語音介面傳遞資訊給使用端。因此，相較於按鍵式服務，語音介面將更易於使用，故此應用將適用於銀行、股票報價與交易之客戶服務系統。透過語音自動化服務，並藉由此自然且直覺式的使用者介面來提供更廣泛的存取機制與功能，亦相對提升語音應用程式的應用價值。

The web-based application by VoiceXML service on the Internet is gradually accepted for human-machine interaction because it provides the speech-enabled functionality and makes the telephone access a reality. However, it is not cost efficient to build voice only stand-alone web implementation and is more reasonable that voice interfaces should be retrofitted to be compatible or collaborated with the existing HTML or XML based web applications. Therefore, this thesis considers that the web site construction should be able to incorporate multiple access modes so that human beings can perceive and interact with either visual or speech response simultaneously. Under this principle, our research develops an integration web based Mandarin dialog system which adopts ASR, TTS, VoiceXML browser, and VoIP technologies to create user friendly interfaces for GUI and VUI. The user can use either traditional telephone line, cellular phone connection, or even VoIP by personal computer to interact with the VoiceXML server. In the mean time, browse the web content from the Internet and access the same document. The implementation system shows excellent performance and can be easily constructed into banks, tourisms, and e-commerce transactions with VoiceXML for wide accessibility.