Previous research on rendering the problem of defining Web wrappers - programs that automatically extract
content from Web pages - practicable has resulted in two major approaches to wrapper generation, that of machine
learning-based wrapping and that of supervised visual wrapper specification. The goal of this project is to work
towards the integration and combination of these two approaches to obtain a framework for defining wrappers that
ideally combines the advantages of both. In such a combined approach, machine learning techniques could
simplify and speed up the visual wrapper specification process by reducing the number of specification steps to be
carried out by the wrapper designer, while learning techniques could be strengthened by new supervised techniques
that only become feasible in a visual specification environment.