摘要
提出了一种结合最新TDT技术、基于增强学习的优先Web环境主题搜索策略,并以此设计主题搜索器系统。该系统通过引入基于领域知识的TDT文本分类技术,大大改进了基于关键字的Naive Bayes模型主题相似性判别的准确性;通过引入基于增强学习的页面评估函数特征化主题Web环境,有效地提高了稀有信息的搜索能力。试验结果表明,该系统具有较高的实用性。
By combining TDT and on - line reinforcement leaming,this paper puts forward a new Web topic search strategy based on Web environment precedence. The strategy results in an intelligent crawler. The accuracy of obtained documents is improved by TDT technique,which is based on domain knowledge. Using a function, based on reinforcement learning, to value Web pages and then to feature Web topic environment, this method works well in promoting the search efficiency on rare information in effect. The experiments show that this system is more effective.
出处
《微机发展》
2005年第8期145-147,共3页
Microcomputer Development
基金
河北省自然科学基金资助项目(F2004000132)
关键词
智能搜索器
TDT
WEB环境
增强学习
领域知识
intelligent erawler
TDT
Web environment
reinforcement learning
domain knowledge
