A Study on Information Acquisition and Extraction from Dynamic Information Sources Based on Knowledge Discovery and Learning

在本研究项目中，我们研究了基于知识发现和机器学习技术的web环境下信息源的信息提取和用户友好的数据操作设施。研究结果可以总结如下：1。从隐藏的网站中提取信息很重要。我们开发了一种从包含文档数据库的隐藏网站中提取新主题内容的方案。我们发展了一些新的信息分析方法。特别是，开发了从文本流中进行主题检测、异常值检测和相关分析的新方法。我们研究了使用分类技术从web中提取信息。特别地，开发了一个利用现有分类法层次结构的网络信息检索系统。我们还开发了一种新的网页信息提取方法，该方法将面向数据库的集中爬行和网页信息提取相结合。研究了分布式环境下信息共享的基本方案。我们开发了一种从包含噪声的事务数据库中进行规则挖掘的知识发现技术。我们开发了一个智能工具，可以根据用户给出的数据操作示例自动生成XML查询。开发了将原始信息源中的二进制数据映射到XML视图的基本方案。我们研究了能够在考虑原始信息源特性的基础上，将从不同信息源提取的各种信息集成在一起的系统架构。

In this research project, we studied information extraction from information sources in the web environment and user-friendly data operation facilities based on knowledge discovery and machine learning techniques. The research results can be summarized as follows.1.Extraction of information from hidden web sites is important. We developed a scheme to extract new topic contents form hidden web sites including document databases.2.We developed some new information analysis methods. In particular, new methods for topic detection from text steams, outlier detection, and correlation analysis were developed.3.We studied information extraction from the web using classification techniques. In particular, a system for web information retrieval was developed which utilizes existing taxonomy hierarchies. We also developed a new web information extraction method which combines database-oriented focused crawling and information extraction from web pages.4.Basic schemes for information sharing in distributed environments were studied.5.We developed a knowledge discovery technique for rule mining from transaction databases containing noise.6.We developed a smart facility which automatically generates XML queries based on data manipulation examples given by a user.7.Basic schemes to map binary data in original information sources into XML views were developed.8. We studied system architecture which can integrate a variety of information extracted from different information sources taking properties of original sources into consideration.