Our Web Search is set up so that if there is no exact match to the
query you entered, a "fuzzy", or "ranked" search may be automatically
entered into. In that case, a graphic bar is displayed next to each hit
to show you how close or how far the hit is from the query you entered,
where those determined to be closest to your query are listed first.
Our ranking algorithm takes into consideration relative word importance,
document content, and adjacency. The output is ordered with the best
match first, followed by the next best, and so on.
Other search technologies refer to this sort of result listing as
Relevancy Ranking, even though such methodologies do not
present the results by any kind of true relevance.
On the other hand, our Ranked Web Search does use a specific
system to rank by relevance, though such a search
would be less preferable than an exact match.
Since the words that match your query in the case of a Ranked search
are relevant to their position within the entire document, the
highlighting of what matched your query will include words throughout
the entire document, rather than just within the sentence or paragraph
you would normally see on an exact match search.
The ranked search we use works as follows:
- An individual term's importance is inversely proportional to its overall
frequency within the database; i.e., less frequent terms are more
important.
- Term match position within a document is taken into consideration;
the closer to the beginning of a document the better.
- Term match order with respect to other terms is taken into
consideration.
- If two matches are of equal rank value then insertion order into
the database will determine which is output first.
Therefore:
- A document which begins with the exact phrase in the exact order
that the user typed as a query will be ranked as the highest match.
- A document containing only one of the terms in a multi-words query,
where the matched term is in half the documents in the database
will be very low ranking.