Authors: C. T. Howard Ho, Eugene Agichtein, Joerg Gerhardt, Vanja Josifovski
Tags: 2003, conceptual modeling
XML is becoming a prevalent format for data exchange. Many XML documents have complex schemas that are not always known, and can vary widely between information sources and applications. In contrast, database applications rely mainly on the flat relational model. We propose a novel, partially supervised approach for extracting user-defined relations from XML documents with unknown schema. The extracted relations can be directly used by an RDBMS, or utilized for information integration or data mining tasks. Our method attempts to automatically capture the lexical and structural features that indicate the relevant portions of the input document, based on a few user-annotated examples. This information can then be used to extract the relation of interest from documents with schemas potentially different from the training examples. We present preliminary experiments showing that our method could be capable of extracting the target relation from XML documents even in the presence of significant variations in the document schemas.Read the full paper here: https://link.springer.com/chapter/10.1007/978-3-540-39597-3_38
