查詢結果分析
來源資料
相關文獻
- 全球資訊網上智慧型圖文檢索引擎的建構
- WWW資訊檢索的新趨勢--欄位檢索
- 架構在WWW與Z39.50上的近似自然語言OPAC檢索系統
- 架構在WWW與Z39.50上的近似自然語言OPAC檢索系統
- GAIS Computer Science Bibliographies Search
- From Document Retrieval to Web Browsing: Some Universal Concerns
- 模糊理論及其在圖書資訊學上之探討
- 語意式旅遊網站服務比價機制之研究--以自助旅行為例
- IEC Station:一個Java Based的嵌入式系統
- Social Dimensions of the Digital Revolution
頁籤選單縮合
題 名 | 全球資訊網上智慧型圖文檢索引擎的建構=Construction of an Intelligent Image-Text Search Engine on WWW |
---|---|
作 者 | 陳鴻文; 江憲坤; 陳孟君; | 書刊名 | 大葉學報 |
卷 期 | 9:1 2000.06[民89.06] |
頁 次 | 頁59-74 |
分類號 | 028.8 |
關鍵詞 | 資訊檢索; 全球資訊網; 模糊邏輯; 多媒體搜尋引擎; Information retrieval; WWW; Fuzzy logic; Multimedia search engine; |
語 文 | 中文(Chinese) |
中文摘要 | 由於全球資訊網(WWW)應用的蓬勃發展,使得網際網路上充斥著各種多媒體資訊,其中不乏以圖形的方式呈現。自然而然地,如何有效的管理及檢索這些圖文資料,就成為當前資訊檢索的重要課題之一。雖然目前WWW上已有少數多媒體搜尋引擎的存在,但這些搜尋引擎由於需以半人工方式處理網頁,且資料更新速度慢,故容易造成資料量過少,使用者需以預設的分類方式瀏覽等缺失,並不足以完全滿足使用者實際的需求。因此,如何快速而又有效地檢索出與主題相關的圖形檔,以大幅節省使用一般搜尋引擎檢索圖檔所浪費的人力、物力,是本研究的主要目的。 本研究嘗試提出一個模糊比對的圖文檢索系統之雛型,主要是由主題知識庫推論模組、委託檢索、資訊攔阻、網頁檢索、圖形檢索與圖形展示,共六個模組所組成。系統運作主要是利用事先建置的主題相關知識庫及辭彙庫,先將使用者輸入的查詢關鍵字,轉換成更完整且具效率的關鍵字群,再委託一般的全文檢索搜尋引擎,找出相關的網頁。之後利用文字比對速度快的特性,嘗試分析出網頁的HTML文字特徵,包括了標題、文件內容及圖檔名稱,再輔以模糊法則推論,來評估出與查詢主題的相關程度,以找出與主題吻合度高的圖形檔。最後以汽車網頁為檢索對象的實驗效率,足以顯示本雛型系統的可行性。 |
英文摘要 | As the rapid growth of applications within the Internet and WWW, graphic and image files are widely used in multimedia webpages. Thus, the way of effectively managing and retrieving webpages with image files has become an important issue today. Although there are a few multimedia search engines available, e.g., Yahoo's image search engine, AltaVista, Infoseek, they cannot completely match users' demands because of imprecise retrieval results. In addition, a manual update of webpages usually makes so slow and small amount of data accessible in those search databases. Therefore, a quick way of retrieving relevant webpages with image files is pursued and reported in the research. There are six modules in the proposed prototype system, including domain-knowledge inference, URL retrieval commitment, URL filtering, webpage retrieval commitment, image-file evaluation and presentation modules. Utilizing the constructed vocabulary and domain-knowledge databases, the search keyword typed by users will be automatically transformed to another group of keywords to make the retrieval process more efficient. Then, related webpages will be reported through full-text search engines and collected by our webpages retrieval module. The HTML features of webpages, including title, keywords within the document and image filenames, are first analyzed by a quick character-string match method. Following that, a fuzzy inference module, which uses scores of those HTML features as inputs, can measure the degree of topic relevance for all image files contained in the collected webpages. Experiments show that the novel image searching technique can illustrate a promising performance in both retrieval time and correctness in processing automobile webpages with images on Internet. |
本系統中英文摘要資訊取自各篇刊載內容。