Information Monitoring System on World Wide Web

摘要

The overall objective of this effort is to explore the use of WebMon, the Web information monitoring system, to conduct social behavior analysis. WebMon was originally created to conduct web monitoring on selected domains, such as the Australian Government Web pages and health news web pages. Information overload is one of the main problems in web monitoring systems, so to overcome this problem the authors developed a web document classification system. The Multiple Classification Ripple-Down Rules (MCRDR) knowledge acquisition method was employed to facilitate incremental knowledge encoding for classification. This project attempted to resolve the technical challenges necessary to enable WebMon to collect data on the influence of Internet-based social networks on social issues and opinions. The following are the two major tasks that were undertaken: (1) enhance the monitoring capability of WebMon so that it can collect data useful for social network analysis, in particular the formation of public opinions; and (2) analyze the data collected and find the factors that affect public opinion formation. The enhancement of WebMon entailed the addition of smart multi-context filtering, smart scheduling, and sentimental context classification. Smart multi-context filtering allows one to distinguish comments sections, or opinion sections, from main content sections. This allows one to study the patterns of comments, including their number and the impact of previous comments to follow-on comments. The smart scheduler helps the system estimate how often new comments are updated. The third enhancement is the use of the MCRDR method to classify opinions, which contain sentimental information and are much more difficult to classify than issues- oriented content.