簡易英文會話系統之建置
摘要
學習外語有助於與人溝通無礙、增加就業及升遷機會、通過考試等，讓人的觸角延伸至世界各地。而眾多的語言種類之中，英語已是全世界的通用語言。隨著地球村的到來，流利的英語能力變得非常重要，對每一個出國留學、旅遊或從事經貿工作的人而言，英語學習更是有許多的優勢。 學習英語並非一蹴可及而是需要時間和精神的付出。目前已有很多免費或付費的課程或網路上的資源可以協助人們提升及改善其在英文寫作、聽力和口語能力。而這些課程有的是集體教學，有的是一對一對談教授英文，也有的是自學書本或線上課程等。學好英語的秘訣就是置身於一個真實的英語生活環境。但有些人在課堂環境中，面對著和自己有相同母語的同學前講英文感到奇怪且不自在。此外，人們已習慣於依自己的步調及能力進行適合的個人學習。 本論文應用現有HTML5網路技術，以Google translate的文字轉語音及語音辨識的API，建置一簡易的英語會話平台。在此平台上學習者可以依自己的需求及步調進行英文聽力及對話的練習。而這些學習的對話內容可以是線上內建的資料或自訂的內容。在發音練習中，本系統以API辨識後的文字進行字串相似度運算，並將計算結果的相似度訊息回饋給學習者，可讓學習者知道哪些發音是可辨識、哪些發音是辨識錯誤。此相似度計算方式是依據本論文所提出的新相似度計算公式。由系統回饋的相似度訊息，學習者可以知道其英語會話能力。此功能將有助於學習者提高他們的自信心和英語會話能力。

關鍵字：英文會話系統、語音辨識、文字轉語音、相似度、HTML5。

The Implement of a Simple English Conversation System
Abstract
Abstract Learning a foreign language can connect one with the world, help him or her to communicate with the local community, increase the chance of finding a job or get a promotion, pass exams, etc. Among so many languages, English is a universal language which has been widely spoken by people in the world. With the advent of the global village, the need of the abilities of understanding English and speaking English fluently become very demanding. No matter who one is (e.g., a student, a business man, a tourist, etc), one may benefit by learning English. Learning English takes time and dedication. There are a lot of free or chargeable classes or web sources which can teach one to improve his or her English writing, listening and speaking capabilities. These courses may be group courses, individual courses, self-study textbooks, or on-line courses, and so on. The secret to learn English well is to put one into a real English living environment. Some people don't feel comfortable in a classroom environment. Especially, one has to speak English in front of classmates whose mother language is the same as him or her. In addition, one may prefer to study at his or her own pace and when it suits him or her. This paper propose a simple English conversation platform based on the use of an existing network technology HTML5 and Google’s text to speech and voice recognition API tools. Via the simple English conversation platform, users can easily practice English conversation dialogues at his or her own pace and when it suits him or her. These conversation contents can be either ready-made or customized. During a pronunciation practice, the system will feedback a similarity degree message to users to inform them how good their pronunciations were after the system has recognized users’ speech content via the speech recognition API and the computation of the similarity between the correct sentence and the sentence spoken by the users. A new similarity measure is proposed in this thesis for the computation of the similarity degree. From the messages provided by the system, users may know how good they are in English conversations. This function will help users increase their self-confidence and ability in English conversations.

Keywords : English conversation system, Voice Recognition, TTS, Similarity, HTML5.