[Sigia-l] String search question

[Christina Wodtke]

I'm not sure how much customization you can do with you search engine, but
I'll assume the best


> 1. This system is not a content site but an content management system
where
> people are going to search the contents and the metadata of unstructured
> data. It is similar to a content site, but different in that the name of
> objects is more important and more similarly behaves like an OS file
system.
>
> I think this is why the vid=video is important.

Only if your users are engineers or IR specialists... I'm not sure this is
something you can assume with a CMS, but I've been wrong before. ;-) Also,
even engineers and researches (like my MR. ) tend to use boxes like they
would a yahoo! search and then use more sophisticated search techniques,
from what I've seen in the lab.

When designing for search, there are two things you wish to avoid: null
searches and irrelevant results. So you want to work your way down form most
likely matches to least likely matches, making sure you always have some
matches (users rarely think it's them when they get no results, they think
your engine is broken or you don't have what they are looking for...)

So a potential schema would be to first search for an exact match for "vid"
then expand to video and then maybe david.

For spring dresses look for "spring dresses" first, then look for "spring"
or "dresses"-- in other words, start with "and" and roll over to "or".
Partial word matches should be a last desperate try. I was searching the
yahoo.

there are a ton of other smart things you can do as well, from adding a CV
behind it to allowing for manual best bets on the head of a zip curve, but I
assume since it's a CMS this is a tough call for you since the content is a
mystery....