Re: suggest: make searches efficient

On 06/01/2008, D Bera <[email protected]> wrote:
> Hey,
>   These days searching gives me headache. There is too many
> information out there. I frequently have to use complicated and
> cleverly thought query syntax to narrow my searches (same for Google).
> I always change my query if I have to visit more than the first 20
> results (again, same for Google). I never realized that personal
> information space could be so vast and varied :-/.
>
>   So I ask you for suggestions, how to make searches efficient ?
>
>   Please suggest only tangible methods e.g. no neural network based
> context-sensitive feedback tracking system to determine what the user
> had in mind. We will leave dashboard for that :)
>
>   E.g. frequently I know what kind of things I am looking for - a
> photo, an email, a contact, an application etc. The query syntax
> allows you to type "filetype:image" or "type:Contact" or
> "source:application" to specify the needs. But thinking and typing
> that elaborate thing will often take enough time to destroy my desire
> in continuing searching. Could be add interesting query syntax to make
> it easier ? It has to be short and easy. But it should not deviate
> from the obvious meaning e.g. if we decide adding "videos" to the
> query would only query videos then that would destroy searches with
> the intent of finding the word "video" in the data/metadata.
>
> Anything that could be done on the UI side ? A drop box of every
> possible data type ? My problem with a drop box is if it is naively
> filled with entries, I have to click on it, scroll through the list
> and then click again to select the type ... thats 2 clicks and a huge
> careful scrolling!
>
> This will require an ongoing effort, so be creative.
>
> Email partially inspired by looking at the deskbar, gnome-do,
> kbeaglebar etc. On small systems, they work nicely and return you the
> apps or contacts etc. But they are designed to encourage you to look
> in the first 5-10 results only. That does not work in a long used
> system with lots of different data. Note that quicksilver or launcher
> specific apps do not have this problem since they already search
> within a narrow search scope.
>

How about: http://www.statsbiblioteket.dk/search/index.jsp?query=beagle&first_search=true&action=classic

Ie cluster/facet extraction on the result set? It is far from trvial
to do in an efficient and scalable way, but it can be done... The
website I linked to have 10M items in the index.

Cheers,
Mikkel