請用此 Handle URI 來引用此文件:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/7184
標題: | 自傳式記憶輔助型對話系統 Dialogue System Assisted by Autobiographical Memory |
作者: | Yu-Ting Hsiao 蕭羽庭 |
指導教授: | 傅立成 |
關鍵字: | 對話系統,自傳式記憶,自然語言理解,核心對話模組, Dialogue System,Autobiographical Memory,Natural Language Understanding,Core Chat, |
出版年 : | 2019 |
學位: | 碩士 |
摘要: | 老人照護是目前人們很關注的一個社會議題。將機器人導入老人照 護的領域有助於降低老人照護的人力需求。其中,社交型機器人能夠 通過與老人的交流為老人帶來心靈上的支持。
在本篇論文中,目標是能夠開發一個能夠和老人聊他們的故事的機 器人對話系統。此系統的關鍵在於將一個能進行科學計算的記憶模型 引入到對話系統中,讓機器人擁有記住人類個人資訊的能力並且能夠 依照相關的個人資訊對人進行回覆。為了了解人類的語言並且從中抽 取出合適的資訊,本篇論文中提出了一個自然語言理解的模組,由基 礎的自然語言處理工具、模板及規則結合而成。在對話管理模組的控 制下,從人類語言中提取的資訊會被送到自傳式記憶模組,一個負責 儲存人類記憶並且能夠抽取相關記憶內容的模組。之後,抽取出的相 關記憶內容會被送到 Skills 模組,生成機器人的回覆。最後,生成的 機器人回覆會經由自然語言生成模組展現給人們。展現機器人的回覆 的同時從記憶模組中抽取出的記憶內容也會被展現出來。 系統的評估經由設計好的幾個實驗。實驗的結果顯示本篇論文提出 的系統能夠利用人類的語言抽取記憶資訊並從記憶的知識庫提取相關 的記憶內容。並且此系統能夠將記憶內容融合到機器人的生成語言中, 以此和人盡心交流。此揭示了機器人能夠對老人提供心理層面照護的 可能性,以及和老人建立陪伴關係的可能性。 Elderly care is a social issue that humans are focus on. It is helpful to introduce robots into the field of elderly care, which alleviates the requirement of manpower of taking care of senior citizens. Social robots can especially provide elders with mental support through communication. In this thesis, the aim is to develop a dialogue system for robots which can chat with elders about their own stories. And the key is to embed a computational memory model into the dialogue system so as to give robots the ability to memorize human’s personal information as well as to generate relevant responses. To understand human utterances and extract appropriate information, a Natural Language Understanding module is proposed, consisting of rules, patterns and typical natural language processing tools. Then, under the control of designed Dialogue Management module, the extracted information are sent to Autobiographical Memory module, where memory of humans is stored and memory items are retrieved. After that, the retrieved memory items are sent to Skills module, from which various robot’s responses are generated. Finally, the generated responses are presented to humans through a Natural Language Generation module, which shows both robot’s utterances and depicts retrieved memory items. The system is evaluated through several well designed experiments. The results show that the system proposed is able to utilize human’s utterances to extract memory information and retrieve memory items from knowledge base. Moreover it is also given the ability to generate responses embedded with memory items to chat with humans, which reveals the potential for offering mental supports to elders so as to build the companionship. |
URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/7184 |
DOI: | 10.6342/NTU201903356 |
全文授權: | 同意授權(全球公開) |
顯示於系所單位: | 電機工程學系 |
文件中的檔案:
檔案 | 大小 | 格式 | |
---|---|---|---|
ntu-108-1.pdf | 3.36 MB | Adobe PDF | 檢視/開啟 |
系統中的文件,除了特別指名其著作權條款之外,均受到著作權保護,並且保留所有的權利。