Thoughts about search interface discussion

Hi team,

Kevin just called me up and we discussed the search interfaces presentation
in a bit more depth. Here's a quick summary of our conversation - Kevin
please let me know if I've got anything wrong - aimed at Mick to help with
the demo script, but presented here because it may be of interest to other
team members. 

Kevin Smathers:

I think a browsing interface is mandatory. Starting from scratch with just
search by keyword is hard if you don't know what the repository contains.

I liked the faceted search interfaces, they seemed to be good browsing
interfaces.

Of course, browsing has to be done in a sensible way. On one system browsing
was alphabetic, and creating categories simply based on alphabetic order
doesn't make that much sense. There must be better ways of classifying /
categorising.

Regarding the query searches, it seemed you could break the query searches
into two subtypes:

- keyword search

- "anchored search". Here the system had some kind of index, so if you typed
in a term that was not in that index, it would position you in the index at
the closest word to the one you just typed. One particular system tried to
do this, but it had problems because it had hyperlinks that didn't work or
at least linked to things that didn't exist in the index. It seemed their
index was out of date with their data. For example if you clicked on Italy,
instead of taking you to Italy in the index it took you to the right place
but Italy wasn't there. 

If we were going to build an index based system we should try to use a
permuted index and try to keep it up to date with the content.

One interesting approach on some systems was instead of presenting a library
card catalog interface, you would get a museum catalogue interface e.g. a
clear order for presenting the work. They didn't present the metadata at
all, just images and information such as a short description of the work.
Its an interesting interface because it brings out things you wouldn't see
if browsing through the catalog, but of course requires human intervention.

Mark Butler:
Yes, we've spoken a bit about communities annotation. When people mention
this, originally I thought they just meant annotations to the card index
information. It's now clear to me that this is not the case, as communities
are very interested in the relationships between items. One of the image
libraries demonstrated this, by providing a view where you could compare
different images and read a commentary on the comparison, another view which
was a tour through a set of images, and another view which placed images in
a geographical context. So it seems we need to leave plenty of flexibility
for different communities to deal with annotation.

Kevin Smathers:
Some interfaces I didn't like. On the National Institute of Health site,
there were lots of links to click one that weren't much use e.g. provided
some metadata that wasn't really aimed at the user and encoded in an arcane
format. If we have admin access like that, we only want to make that
available to the correct people e.g. via login or expert user.

There were variations of anchored search for collections with multiple
indices, where once you typed in a keyword the system decided which of the
indices you were trying to search. This was a little strange at first (until
you realised what it was doing).

Mark: 

So it sounds like we are in agreement on a number of points then:
- For the demo, we need some form of browse as well as some form of query.
- We need to control what metadata we make available to the user, as
different users have different metadata requirements.
- There is scope for work here on community annotation, but that's beyond
the scope of the initial demo. 

Dr Mark H. Butler
Research Scientist                HP Labs Bristol
mark-h_butler@hp.com
Internet: http://www-uk.hpl.hp.com/people/marbut/

Received on Wednesday, 10 September 2003 10:40:12 UTC