Automated synthesis of frequent event-sequences corpus from large-scale textual data and its application to WEB content tracking

点击翻译按钮获取中文摘要

This research we studied and developed the following technologies:1. a novel and rational frequency measure, called Total Frequency Measure, which satisfies anti-monotonic property and never causes duplicated counting within a very long single data sequence.2. a online fast sequential data mining algorithm for extracting frequent subsequences within the framework of a infinite-length window.3. a fast sequential mining algorithm based on the relaxation method which is intended for use for the framework of a finite-length window.4. a intelligent sequential data mining method which uses an integrated occurrence criteria of frequency and information gain for subsequences.5. a sequential pattern mining method for WEB access logs, which enables us to analyze access log data with considering page-staying time sequences6. a new method for extracting important key words and/or phrases from newspaper articles in a huge newspaper corpus.We showed the significance of the above technologies throughout huge amounts of experiments for evaluation.