Title: | 在搜尋引擎內利用連結關係完成使用者設定描述 User Profiling by Link Relationship in a Search Engine |
Authors: | 鄭國宏 Kuo-Hung Cheng 鍾崇斌 Chung-Ping Chung 資訊科學與工程研究所 |
Keywords: | 連結關係;使用者設定;加權值;連結值;link relationship;user profiling;authority value;hub value |
Issue Date: | 2001 |
Abstract: | 隨著網際網路的資訊越來越豐富,使用者必須花更多時間來尋找其所需要的資料。搜尋引擎是個在網際網路中擷取資訊很有用的工具,它根據使用者所查詢的字來做回應。但是在真實狀況下,對於相同的查詢字而言,每一位使用者所想要的結果是不同的。
本篇論文的目的,是要對網路上查詢的結果做個人化。我們提出一個利用連結關係來做使用者設定的方法。和向量空間模式不同的是,我們所提出的方法不依靠文字而只依靠在網際網路中網頁上的連結關係。利用這個連結的關係來衡量網頁使用者設定的加權值與連結值,並用這些值來做網頁的排序,而且事先預測使用者喜愛的網頁。簡而言之,如果一個網頁被很多其他網頁所連結,我們則認為這個網頁是有價值的。進一步來說,如果一個網頁被很多同一使用者曾經選過的網頁所連結,我們則認為這個網頁也應該是此使用者所感興趣的。在我們的模擬中,利用現在網路中實際的網頁連結關係與各種不同的預測使用者興趣來評估預測的準確率,此結果似乎是可被證實的。 As information in World Wide Web becomes richer and richer, the users have to spend more and more time finding what they want. The search engine is one of the useful tools to retrieve information in WWW and it replies to users according to the query terms. But the real status is that user wanted results of the same query term are different respectively. The objective of the thesis is to personalize the web query result. We proposed a method by using link relationship to do the personalization. Differing from the vector-space model, the proposed method depends not on terms but only on link relationship between pages in WWW. The link relationship would be used to evaluate the personal authority value and hub value. These values are also used to rank pages and speculate the user favorite pages in advance. In brief, if many other pages link a page, we think this page should be valuable. Furthermore, if a particular page is linked by many others pages which the particular user chosen before, we think this page should also be interesting to the user. In our simulation, we use the link relationship between pages in World Wide Web and different user’s interests to evaluate the speculation. The results seem to be promising. |
URI: | http://140.113.39.130/cdrfb3/record/nctu/#NT900392069 http://hdl.handle.net/11536/68480 |
Appears in Collections: | Thesis |