Wrapper generation

• Issues
  - refer to the section on Data Models and Schema Integration
  - Mediator systems usually require more complex wrappers than do most warehouse systems
• Ways of creating wrappers
  • Manual
    Why is it impractical for some sources?
    In case of Web sources:
    - big number of sources
    - new sources are added frequently
    - format of sources change
    So, high maintenance costs.
  • Semi-automatic (interactive)
    Noted that only small part of the code deals with the specific access details of the source. The rest is common among wrappers or data transformation can be expressed in a declarative fashion (high-level). Graphical interface, programming by demonstration.
  • Automatic
    • site-specific or generic
    • usually needs training often supervised learning
• Tools for semi-automatic/automatic wrapper construction for structured/semistructured data
  • Template-based wrappers
  • Inductive learning techniques for automatically learning a wrapper (using labeled data)
    Inductive learning - task of computing some generalization from a set of examples
    Methods:
    • zero-order (decision tree learners)
    • first-order (inductive logic programming)
      - bottom-up/top-down approaches
• Tools for data extraction from unstructured documents
  • Using ontologies and conceptual models to extract and structure information from data-rich, unstructured documents.
  • Using heuristic approaches to find record boundaries in web documents.