Disambiguation Web Appearance of People by Exploiting Link-based and Content-based Information of Social Network Relation

• Main Authors: Kun-Yan Chiou, 邱坤彥
• Other Authors: Hahn-Ming Lee
• Format: Others
• Language: en_US
• Published: 2007
• Online Access: http://ndltd.ncl.edu.tw/handle/s48u79

碩士 === 國立臺灣科技大學 === 資訊工程系 === 95 === Searching for personal data about people of interest is one of the most popular types of search activity. However, with the exponential growth of the WWW, users need to spend much time on filtering none interesting web pages. This is because personal names are not unique, many different individuals with same person name are found in search result web pages. Currently, many search engines provide functions for automatic classifying web pages; however, there is less function for personal name classifying. Therefore we try to provide a system which can automatically group web pages referred to different individual. The web pages grouping result will reduce the time consuming of users; they can effectively browse these public web pages referred to target individual. Many past researches utilized predefined template rule or traditional content VSM model to measure the similarity between different web pages, but these kinds of methods have some limitation in real situation. In this thesis, we propose a link-based method by exploiting social network relation in the WWW, then grouping together searched web pages referred to different individual. Additionally, content-based information adopts collect part of web pages which have fewer link relations. Our proposed method need no additional background knowledge for target personal name, thus user can reduce the confusion on setting additional query keyword. The experiments result show that proposed methods can get good achievement in general.