使用基因演算法與相關回饋於協助網頁搜尋; Using a Genetic Algorithm and Relevance Feedback in Assisting Web Search

NCUIR > School of Management at National Central University > Graduate Institute of Information Management > Electronic Thesis & Dissertation > Item 987654321/12935

Please use this identifier to cite or link to this item: http://ir.lib.ncu.edu.tw/handle/987654321/12935

Title:	使用基因演算法與相關回饋於協助網頁搜尋;Using a Genetic Algorithm and Relevance Feedback in Assisting Web Search
Authors:	張永霖;Yeong-Lin Chang
Contributors:	資訊管理研究所
Keywords:	基因演算法;相關回饋;資訊檢索;詞頻;全球資訊網;relevance feedback;genetic algorithm;World Wide Web;information retrieval;term frequency
Date:	2002-06-21
Issue Date:	2009-09-22 15:19:15 (UTC+8)
Publisher:	國立中央大學圖書館
Abstract:	全球資訊網的資料量從發展之初至今呈現加速度成長的趨勢，早已成為人們獲取資訊的重要管道之一，因為其資料量龐大，全球資訊網的使用者遭遇了資訊過載的問題，因而求助於目錄服務與搜尋引擎。但目錄服務限於維護人力不足，提供索引的網頁往往不敷使用；搜尋引擎的搜尋結果與使用者的資訊需求往往相關性過低，或搜尋結果過多，使用者仍須自行過濾所需的資訊。本研究試圖建置一個架構於現有搜尋引擎之上的智慧型代理人。根據使用者提供相關的範例文件，建立使用者興趣檔。運用基因演算法搜尋可能的查詢字串，透過搜尋引擎蒐集網路上的相關網頁，以向量空間模式表示各網頁文件的內容，並評估網頁與使用者興趣檔的相似程度，藉此引導基因演算法搜尋更適合的查詢字串。並根據使用者對檢索結果的評等，配合相關回饋機制調整使用者興趣檔，逐次改進查詢的效果。本研究實作之系統兼顧無檢索主題限制、軟硬體需求低且使用者額外負擔少等方面，對於網頁搜尋及使用者興趣學習上，有令人滿意的表現。 World Wide Web(WWW) is growing faster and faster since its emergence, and is one of our major information sources in daily life. Because the data quantity of WWW approximates to infinity, information-overloading problems bother users all the time. To retrieve information in the right scope and content has become an important issue. Presently, directory services and keywords searching are two major ways that can help users. Unfortunately, both ways have shortcomings. Referring to directory services, the problem is that a labor-intensive activity is required to create and maintain directory. Besides, web pages are usually not fully indexed. As for keywords searching, the problem is that too many unrelated information are usually provided that users have to spend a lot of effort on filtering. Our research tries to construct an intelligent agent to assist information retrieving. It builds a user profile by parsing and analyzing example documents provided by the user. It uses a genetic algorithm to search the possible query strings combined by the keywords from the user profile. It collects the relevant web pages via search engine. Each web page is represented in vector space model. It tries to search more fitting query strings by evaluating the similarity between web pages and user profile. According to the user evaluations, relevance feedback mechanism refines the user profile to improve the query results. This proposed system provides a satisfying performance in web search and learning users’ interests. It can work for every search subject with low software and hardware requirement and less user extra interferences.
Appears in Collections:	[Graduate Institute of Information Management] Electronic Thesis & Dissertation

Files in This Item:

File	Size	Format

社群 sharing

Loading...