I3S - An Intelligent Internet Information System for Constructing Domain-Related Knowledge Portals

[[abstract]]在現今網路發達的時代，使用者依賴網際網路取得大量相關資訊，以供參考，並尋求解決問題的策略。使用者可以從搜尋引擎中找到許多相關資訊的網頁，仔細閱讀並分析這些網頁資訊，才可能整合出他們所需要的「有用知識」。網際網路可以做為解決各種不同問題的工具，但目前還沒有一個以Web為基礎的智慧型系統，可以自動地處理各種不同領域的問題。因此，我們開發了一套智慧型的網際網路資訊系統 (I3S)，用於建立不同領域的知識入口網站。I3S整合了Data Mining的方法和資訊擷取的技術，可以從網頁找出領域相關的概念，並擷取出metadata整合到我們的知識入口網站，供使用者參考。在經營不同領域的知識入口網站時，會有不同的metadata需要整合，不同類型資料也需要不同呈現的版面。為整合跨領域metadata，我們將領域metadata的欄位對應到數位圖書館共通的標準，Dublin Core所定義的十五個欄位。以XML做為資料的輸出，透過XSLT將XML轉換成為HTML來呈現各領域入口網站不同的樣貌。透過Web-based介面，領域專家可與I3S互動，以有效地以半自動化的方式建立領域入口網站的雛形 (例如：旅遊)。為了因應不同領域需求的不同資料模式和管理流程，本系統也利用資料庫管理系統所提供的檢視 (View)、XML表達方式和搭配不同的XSL來做為版面的呈現，讓I3S的系統管理者可以針對不同領域的應用，客製化建構和管理領域知識入口網站。從所花費的時間與資訊正確性的角度評估，以旅遊領域為例，領域專家僅需花費和系統互動不到一小時的時間，就可以從擷取後的資料中建立此領域的概念和知識入口。因此，I3S可以做為一個解決領域相關問題的網際網路智慧型資訊系統平台。[[abstract]]In the Web era, users rely on the Internet environment to find information needs for making decisions or solving problems. Nowadays, users can easily obtain millions relevant pages from search engines. However, they must perform tedious “click-and-click” operations to elicit useful information from numerous pages. The key problem is that users require not only “relevant pages” retrieved by search engines, but also “useful knowledge” integrated to make decisions. Undoubtedly, the Web can be regarded as a potential infrastructure for solving various domain-related problems. However, no integrated intelligent systems are proposed to automatically explore the Web for handling domain-driven problems. In this thesis, we develop an Intelligent Internet Information System (I3S) as a general platform for constructing knowledge portals for various domains. By integrating several data mining methods and information techniques, I3S is able to recommend domain-related websites, recognize significant structured domain concepts as the basic ontology, extract domain metadata from pages, and integrate all of them into a domain knowledge portal to support decision makings. Different domains have different metadata presented with various layouts, thus the customization problem of constructing various domain portals is also considered in the design of I3S. Taking XML as the common data format, I3S is customizable by providing templates in XSL files that transform XML into different HTML presentations for different domain portals. By mapping different domain metadata fields into 15 Dublin Core elements, I3S also achieves the metadata integration of various domains. With few interactions with domain experts, I3S semi-automatically constructs a prototype for the “Travel” domain as the case study. Evaluations on effectiveness and time-cost demonstrate that I3S can be extended as a general web platform for solving domain-related problems.[[note]]碩