回首頁校友資訊相關網站搜尋引擎∣檔案下載

 

 

研 究 生:張明華

指導教授:黃永廣  博士    Wong, Wing-Kwong

學 年 度:88

學位類別:碩士

論文名稱:幾何知識庫與自然語言代理人

 

論文摘要:

在本篇論文中,我們提出一個以幾何知識庫為基礎的自然語言代理人架構,並以此架構為
藍本發展出一個『幾何知識自然語言代理人系統』(Natural Language Geometry
Knowledge Agent )來輔助學生學習關於幾何方面的知識。
透過這個幾何知識自然語言代理人系統,學生可以用自然語言來描述他的幾何問題,再由
『幾何知識自然語言代理人系統』跟『COOL/CLogo網路繪圖學習環境』溝通,透過系統所
產生的CLogo程式,以及在網頁上所顯示的幾何圖形,來進一步引導學生學習有關幾何方面
的知識,同時也可以建立學生對於程式設計的基本觀念。
在這個系統中,學生輸入關於幾何的中文自然語言語句,系統便能理解出此語句真正的語
意,並從題意中找出幾何條件,再根據這些條件來輸出對應的中文物件導向描述語言(
COOL Script),傳送給中文物件導向描述語言伺服器(COOL Server) 以產生題意中隱含的
幾何條件,同時將此描述語言(COOL Script)編譯成標準的CLogo指令,由CLogo繪圖學習環
境來繪出學生解題所需要的幾何圖形,在解題的過程中,學生從隱含許多條件的題意裡,
常常無法找出需要的幾何關係[1],但透過幾何自然語言代理人系統所產生的幾何圖形,可
以從輔助的圖形上提供學生解題的資訊,如此便可引導學生漸進式的學習幾何。
In this paper, we propose a natural language agent architecture based on a geometry knowledge base,
and developed a Natural Language Geometry Knowledge  Agent system to assist students learn
knowledge in geometry. Through the natural language understanding mechanism in the system,
students can describe their geometric questions with natural language to communicate with CLogo
Network Learning Environment for drawing the figures they needed to solve their geometric problems.
With the CLogo programs generated by the system and he geometric figures displayed on the CLogo
Network Learning Environment, students are guided to organize their notions about geometry and to
learn the essential concepts about computer programming language. The semantici nformation of the
natural language descriptions about the geometric question that student input is grasp and is processed
by our system to find out conditions of the geometric figures for generating standard COOL Scripts.
Furthermore, those COOL Scripts can be compiled into CLogo programs by COOL Server for
drawing geometric figures on the CLogo Network Learning Environment. As students attempt to solve
a geometric question, they may ignore the implicit conditions involved in the question. Through the
geometric figures shown in CLogo Network Learning Environment, the information assisted to
solve the question are provided. Therefore, students can learn geometry at ease from our
geometry-based agent system.