Hierarchical Concept Description and Learning for Information Extraction

This paper addresses the problem of extracting information from textual documents, either normal documents or web pages. A new approach for extracting complicate information from semi-structured documents is introduced that exploits a successive hierarchical rule-learning algorithm. Through evaluation it is shown that this approach can extract complicate concepts with a much higher precision than the equivalent rule learning applied to flat text. In addition, the rate of learning is significantly higher for the hierarchical approach.