Extracting Network Preferential Summary with Bootstrapping Method

• Main Authors: Yan-FuCheng, 程彥輔
• Other Authors: Hei-Chia Wang
• Format: Others
• Language: zh-TW
• Published: 2015
• Online Access: http://ndltd.ncl.edu.tw/handle/55868538860016135710

碩士 === 國立成功大學 === 資訊管理研究所 === 103 === The output value of e-commerce has obviously growing in 2008. Consumers have most interest in discount and preferential information. It’s difficult for search engine to keep latest and the most comprehensive search result. This research use bootstrapping method with text mining. After determine preferential keyword, set the website that has complete preferential information as seed pages. Finding document object model (DOM) position of preferential information with XML path language (XPath) to get the pattern that can extract preferential information. The pattern will download webpages from chosen websites. Analyzing these pages with word segmentation system and Distance Point-Wise Mutual Information (DPMI), learning new preferential keywords with bootstrapping method. Combine preferential keyword and store or product name for search engine to find out new preferential websites. Developing a user interface which provides preferential information like: buy one get one, buy one, get one half price, etc. Experiment result shows that DPMI using two as word distance has the greatest precision 29.4%, 9.4% higher than PMI’s result 20%.