A technique that is gaining popularity is the meta-engine. A meta-engine is simply an engine that submits a query in parallel to distinct sub-engines and collates the results somehow. This has been done successfully using both traditional IR systems [2, , ] as well as Web-based systems [47, ].
The meta-engine technique is designed to return more relevant documents on the initial query. However, the technique may be applicable to term weighting and selection techniques as well. The assumption here is that terms weighted highly by multiple techniques are more important than those weighted highly by a few or one; likewise, terms selected by multiple techniques are likely more important that those selected by one.
Thus, the feedback system envisioned here would be a combination of selecting those terms that several term selection formulas used, and then weighting those terms using some combination of weighting formulas. Based on results from other fusion methods, I would hypothesize that this method would be an improvement over existing methods, but it is an open question as to how well this approach would actually perform.