TopX is a highly efficient and effective search engine for XML data that
is used, for example, for several tracks of this year's INEX benchmark.
However, for some diffcult queries, the results provided by TopX are not
yet completely satisfying. Towards the solution of this problem, an
extensible framework has been proposed that incorporates feedback from the
user to generate a better, expanded query.
This framework has been developed independently of TopX and is not
integrated with TopX' user interface. The effectiveness of these
approaches has been demonstrated with benchmarks, not with real users. On
the other hand, the efficiency of the implementation has never been an
issue so far.
The goal of this thesis is a tight integration of the feedback framework
and the TopX search engine. This includes several aspects:
1. the extension of the existing, browser-based interface to support
explicit relevance feedback (and, as an optional extension, also
implicit feedback derived from the user's actions), reevaluating the
query when new feedback is available.
2. the modification of the existing TopX engine to support incremental
expansion of queries, i.e., if a query that is already evaluated is
expanded (based on feedback), the evaluation of the expanded query
should reuse the partial results from the evaluation of the original
query.
3. optionally, exploration and evaluation of structural feedback methods.