從網際網路搜尋與瀏覽資訊是近年來學術界與實務界積極研究的重要議題。目前關於搜尋與瀏覽機制的研究,除了基本的關鍵字比對之外,更結合個人化分類、點選流向或協同推薦等機制為來辨識使用者意圖與描述使用者喜好,但並未將分類納入個人化的機制中;本研究提出一個主題分類與點選流向分析之分類機制,以網路搜尋的從屬端應用為目標,將搜尋結果進行個人化分類後呈現給使用者,並自動地學習使用者喜好的改變。 為了建立使用者模型,本研究使用中文旅遊網站之文字資料建立旅遊資訊的主題分類樹,同時利用分類樹的五個屬性(節點、節點權重、連結權重、詞彙與詞彙權重)來代表使用者的喜好。當使用者查詢時,系統會紀錄使用者的瀏覽行為;並透過點選流向分析,將使用紀錄分析後更新使用者側寫檔,以學習使用者喜好的改變。 研究結果顯示,個人化分類的機制可以提高瀏覽搜尋結果的效率,也能增加使用者在網路搜尋時的滿意度;表示在一般非專業資訊的網路搜尋中,將查詢結果依照個人喜好進行分類之後,再呈現給使用者瀏覽,可以增加使用者瀏覽的效率與滿意度。
Searching and browsing on the Internet are both important issues in the domain of information retrieval, and there is also high demand for personalization of web search in nowadays. Recent researches of personal web search use many techniques such as personal categories, click-through analyses or collaborative recommendation to identify user intentions and to describe user interests, but there are few researches combines the personalization and classification to build an adaptive personal web search system. In this study I proposed a conceptual architecture of Personal Classification of Web Search based on subject taxonomy tree and click-through analyses in order to improve the browsing efficiency and user satisfaction. In order to construct user profile, a hierarchal subject taxonomy tree of travel information was built according to literature review. This tree has five attributes which represent the interests of a single user. Each user has his/her own profile for generating personal categories while searching. The system then adjusts user profiles according to each user’s browsing behavior in order to learn different interests of different users. Textual data in Chinese travel web sites are used for experimental data and a prototype system is implemented in order to evaluate the proposed architecture. The result shows that personal classification is capable of the improvement of browsing efficiency and user satisfaction on web search.