標題: 一個擴增實境建構的問答系統
An Augmented Reality based Question Answering System
作者: 俞恩慈
Yu, En-Tzu
羅濟群
陳志華
資訊管理研究所
關鍵字: 擴增實境;問答系統;類神經網路分類法;Augmented reality;question answering system;artificial neural network
公開日期: 2015
摘要: 近年來,希望能夠自動獲取資訊的需求日趨增加,許多研究致力於討論與發展問答系統,經由互動式問答,問答系統能夠根據使用者需求回覆正確答案。擴增實境為一影像技術,能將虛擬的資料與實際的物件或地理影像結合,是提供互動式使用者經驗最好的工具,因此,問答系統與擴增實境的結合,是未來必然的趨勢。本研究設計了一個問答系統,此問答系統包含了五個部份:斷詞切字、向量空間模型、本體論、詞頻-逆向文件頻率以及資料探勘分類技術,本研究會比較不同的資料探勘分類技術,在效能與效率的考量下,選出問答系統最適合的分類法。根據設計的問答系統,此研究在行動裝置上開發了一個「擴增實境問答系統」,此系統包含了兩個部分:行動裝置端以及雲端伺服器端,行動裝置端能夠支援擴增實境技術;雲端伺服器端則提供了問答系統用於分析使用者需求的計算能力。根據實驗結果,擴增實境問答系統的分類法在效率與效能的考量下,類神經網路分類法最符合需求。所以,「擴增實境問答系統」能夠提供互動式的使用者經驗並有效率且快速地回覆使用者需求。
In recent years, question answering systems (QAS) are emerged due to the driving needs of automatically navigating information. QAS has the ability to automatically respond to users according to their requirements, which can offer good interactive user experiences. Augmented reality (AR), an imaging technology that overlaps data produced by computers onto physical objects or geographical locations in the real world, is the best tool for providing excellent interactive user experiences. Therefore, the combination of QAS and AR is the popular and inevitable trend in the future. This thesis designs a question answering system which includes five modules: segmentation, vector space model (VSM), ontology, term frequency-inverse document frequency (TF-IDF) and classification. Different classification methods are compared with respect to their effectiveness and efficiency. The best one will be selected and used in the proposed QAS. Based on the proposed QAS, an augmented reality question answering system (AR-QAS) on mobile devices is implemented. The implemented AR-QAS includes mobile clients and a cloud server. The mobile clients support AR technology while the cloud server provides the computation ability for analyzing input requirements of the QAS. According to the experiment, ANN is chosen to be the classification method used in the proposed AR-QAS. In consequence, the designed AR-QAS provides an innovative combination of AR and QAS which offers more interactive and user-friendly experiences.
URI: http://140.113.39.130/cdrfb3/record/nctu/#GT070153414
http://hdl.handle.net/11536/125637
Appears in Collections:Thesis