| 摘要: | 社群資源分享系統是目前Web 2.0 平台上非常受歡迎的一種應用，它允許使用者上 傳他們的資源(圖片、影音、註解、書籍、文章等) ，並且自由的加註任何標籤(Tags)在 資源上。雖然這些Web2.0 網站提供了一個好的社群平台，然而目前仍缺乏智慧型搜尋 的機制，來幫助使用者在廣大的共享資源中，有效率的找到符合使用者需求的資源。 如何了解使用者的需求，提供友善的搜尋服務模式，讓使用者充分運用Web 2.0 網 站的資源，是Web 2.0 應用成功的重要關鍵。本研究將以旅遊共享資訊為研究對象，嘗 試提出一個結合使用者意圖(Intension)、領域知識本體(Ontology)、與大眾分類法 (Folksonomy)的智慧型知識搜尋與個人化推薦機制，有效的輔助使用者清楚地表達其搜 尋目的與需求，以進行有效率的資訊搜尋。使用者以自然語言輸入問題描述，同時明確 表達其意圖。藉由意圖的指定，將使得從自然語言的分析中，自動擷取出關鍵字變得較 容易，而且方便進行階段性的搜尋。領域知識本體與標籤的Folksonomy，則進一步輔助 推展出相關的標籤關鍵字。之後，將這些標籤關鍵字組置於Web2.0 搜尋引擎上進行搜 尋，以得到滿足個人需求的資訊服務。另一方面，本研究利用協同過濾方法推薦符合個 人偏好行為的旅遊資訊，利用內容過濾方法推薦具有高度類似特性之文章，以便適時提 供對使用者有用的潛在資訊，幫助使用者充分運用Web 2.0 網站的共享資源。 本研究將採用YAHOO 旅遊網頁做為實驗資料來源。我們將撰寫撈取資料的程式， 以擷取實驗所需資料，並且建立旅遊知識本體以及大眾分類等資料庫。最後，我們以 Google AJAX Search API 建置此系統，進行實驗的分析與驗證，探討結合使用者意圖、 領域知識本體、與Folksonomy 在搜尋與推薦上所能帶來的效益。本計畫所提出的方法 可實際應用在各式社群資源分享網站中，如Del.icio.us 推薦書籤、Flickr 推薦相片、 YouTube 推薦影片、部落格入口網推薦部落格、Amazon 推薦商品、企業內部運用標籤 來標記文件的知識管理系統等，應用將十分廣泛。
Social resource sharing systems are web-based systems that allow users to upload their resources (such as pictures, audio and video, books, documents etc.), and to label them with arbitrary words, so-called tags. Although these Web2.0 websites provide a good social platform, however, intelligent searching mechanisms are still lacked to help the users effectively finding their desired resources. How to provide the friendly search services, let the users fully utilize the shared resources of the Web 2.0 websites, will be the successful key of Web 2.0 applications. This study attempts to integrate the user intension, ontology, and folksonomy to propose intelligent searching and personalized recommendation mechanisms by which the user searching goals can be described effectively. The user input a query question in the natural language, simultaneously expresses its intention explicitly. Because of intention assignment, using natural language processing to pick up the key tags automatically will become easier. The domain ontology and the tags folksonomy are further auxiliary to derive the related tag keywords. Afterward, these key tags are put into the search engine to retrieve the user desired information. On the other hand, this research utilizes the collaborative-based filtering to recommend the suitable information for the user preference, and utilizes the content-based filtering to recommend the highly similar resources, in order to help the user fully utilizing the sharing resources. This study will examine the system performance using the dataset from the YAHOO traveling website. We will establish this system by Google AJAX Search API to verify the effectiveness and feasibility of our approach, and conduct experiments to investigate the searching profits based on the user intention, the domain ontology, and the Folksonomy, respectively. The achievements of the project will be an excellent foundation for advanced applications and researches in many Web2.0 social resource sharing systems, such as the shared photo gallery Flicker, the bookmarking system del.icio.us and so on. The applications are very widespread. |
