(Science Environment for Ecological Knowledge) XML Classification Using Content and Structure

The number of XML documents on the Web has seen a phenomenal increase in the past few years. However, most existing efforts for XML document classification employ standard information retrieval methods based on the document contents alone or the document structure alone. In this paper, we present a system that classifies XML documents based on both their content and structure. In particular, we classify XML documents by a weighted combination of field-wise content similarities and show that this approach outperforms classification that ignores the structure. We also show that weighting the fields differentially outperforms an approach in which each field contributes equally to the classification process. We then present an algorithm that learns the relative importance of each field in order to automatically infer appropriate field weights. 1