We propose a method for modeling complex Web sources that have active user interaction requirements. Here "active" refers to the fact that certain information in these sources is only reachable through interactions like filling out forms or clicking on image maps. Typically, the former interaction can be automated by wrapper software (e.g., using parameterized urls or post commands) while the latter coannot and thus requires explicit user interaction. We propose a modeling technique for such interactive Web sources and the information they export, based on so-called interaction diagrams. The nodes of an interaction diagram model sources and their exported information, whereas edges model transitions and their interactions. The paths of a diagram correspond to sequences of interactions and allow to derive the various query capabilities of the source. Based on these, one can determine which queries are supported by a source and derive query plans with minimal user interaction. This technique can be used offline to support design and implementation of wrappers, or at runtime when the mediator generates query plans against such sources.
展开▼
