There have been some questions about the way that the the ranker works in
the Webinator.
The first question has to do with the number of results returned from a
ranking query. We limit the number of results from a ranked query to 50,
as more results would either have a low relevance, or are a sign of a
vague query.
Several people have asked about "seeding" the documents, so that particular
documents will occur first in the results. If two documents are determined
to have equal rank they will be displayed in the order they are in the
database, so you should walk the important documents first. The ranking
algorithm in the Webinator does not use frequency within a document in
determining the rank, so repeating a phrase many times will not help. This
also prevents people from "spoofing" the engine into returning their
page first.
The ranking algorithm does use proximity in calculating rank, and when
searching for several terms the whitespace, as well as the morpheme
processing may cause only 9 out of the 10 bars to light up, even if it
appears that the exact phrase is there.
John Turnbull