交談機器人之創新研究

Abstract

拜人工智慧技術的發展所賜,近年交談機器人深受大眾喜愛,如 Amazon Echo、 Apple Siri、Microsoft Cortana、Google Assistant 等皆是知名的語音交談機器人。而 Amazon 於今年初向第三方開發商開放其語音智慧助手系統 Alexa 的底層技術後, 更可望迅速擴大語音交談技術於各行各業的應用。有鑒於此,本研究將針對交談機 器人的現況,提出一創新交談機器人—「親屬交談機器人」,期望能讓機器人模擬 出過世親人的模樣,與在世者進行互動,達到陪伴的效果。 本研究的研究邏輯為:透過理解現有交談機器人的運作架構及人類交談的元 素,描繪出新交談機器人可行的運作架構。首先,本研究於文獻探討中先對目前的 語音交談機器人相關技術進行認識及瞭解。接著,透過整合這些相關技術,繪製出 現有交談機器人完整的系統運作架構圖。再來,整理出人與人之間對談時所包含的 元素(交談表現、人格、生理、環境)以及這些元素會透過哪些數位方式(照片、 影片、文字、語音)來呈現。透過這種具邏輯的整理方式來清楚理解怎麼樣才能讓 人類使用者與「親屬交談機器人」對話時,有「在和親人對話的感覺」、以及可以 藉由哪些資料去達到這項目標。最後,將這些資料及元素納入既有交談機器人系統 當中,並進行流程上的修改,繪製出「親屬交談機器人」的可行運作架構。 本研究旨在提出清晰的交談機器人願景,並以具邏輯的方式描繪出達到此願 景的可行方法,可望作為後人的參考藍圖,設計出更具有人性化及個人化的未來 交談機器人。

Thanks to the development of Artificial Intelligence (AI), chatbots are in common use all over the world in this era. For instance, Amazon Echo, Apple Siri, Microsoft Cortana and Google Assistant are popular chatbots which have been widely used in recent years. In this April, Amazon started to open up the technology of its artificial- intelligence assistant, Alexa, to other hardware makers for free. Hence, an applying- scope expansion of voice-controlled technology can be highly expected. In view of this, this thesis is aimed to conduct a study on chatbot and also give an innovative idea — “Relatives Chatbot”. The “Relatives Chatbot” is able to imitate the personality and the voice of the family members who passed away, in order to interact with alive people and keep them company. The thesis is focused on four parts: First, presenting the vision of new chatbot — “Relatives Chatbot”. Second, understanding established chatbot by integrating the voice-controlled technologies and drawing the system model diagram. Third, comprehending the essential factors included in people’s conversations and where to gain these elements. Forth, drafting the feasible operating system diagram of “Relatives Chatbot” by putting the elements into the established operation flow and making some revisions to the flow. Through this arrangement, the factors that makes human-users feel like talking to their family members while talking to “Relatives Chatbot” can be considered. In the end, the system model of “Relatives Chatbot” is done.