There are ways to automatically rank the quality of a query. High quality queries shouldn't be fuzzified. Low quality queries should receive assistance.
Part of the problem is that the query recommenders are overeager. Whatever you type, they try to second guess you and "improve" your query. In a paper we showed that if the answer set to the original query is good enough the search engine should back off and past it unfiltered, whereas if the result set is not very good quality an alternate set of answers in a separate column is often useful.
The same is said to be the case with clippy. The prototype version rarely appeared and when it did it was almost always correct and helpful. The shipping version was this animated character always sitting on the screen, raising the eyebrows in a distracting fashion, and obnoxiously autocorrecting a[i] into a[I].
There are ways to automatically rank the quality of a query. High quality queries shouldn't be fuzzified. Low quality queries should receive assistance.
The same is said to be the case with clippy. The prototype version rarely appeared and when it did it was almost always correct and helpful. The shipping version was this animated character always sitting on the screen, raising the eyebrows in a distracting fashion, and obnoxiously autocorrecting a[i] into a[I].