(Snip)
In conclusion, we should have a search interface containing three basic elements: - a big search box -- Google has a simple front page with a centralised search box for a reason - some common options, like those proposed by FT2:
On 31.10.2008 13:13:13, FT2 wrote:
- Articles
- Articles and article discussion
- Articles, discussion, project space, and project talk
- Project and project talk spaces
- Images, categories, templates and their talk pages
- All namespaces
- Custom
- and finally a "Go" button.
Leon
People are familiar with Google. A somewhat similar interface approach would seem familiar. With an "Advanced search" button.
I'd consider also using a second trick Google does. It tries to pick out a few specific useful links and highlight them first. In our case, pages with <text> in the title, or sequentially in the text as a string, may be more likely to be high quality hits, than pages that "just had the words scattered somewhere in the text". Maybe highlight a few of those? So the output might be:
<Hit #1 + sample text> <Hit #2 + sample text> <Hit #3 + sample text> ---> Click here for more pages with X in the title <Hit #4 + sample text> <Hit #5 + sample text> <Hit #6 + sample text> <Hit #7 + sample text> --> Click here for more pages containing the text "X" <Regular hits from here on>
FT2
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
FT2 wrote:
I'd consider also using a second trick Google does. It tries to pick out a few specific useful links and highlight them first. In our case, pages with <text> in the title, or sequentially in the text as a string, may be more likely to be high quality hits, than pages that "just had the words scattered somewhere in the text".
That's what the ranking algorithm is for -- title matches are already more highly-ranked.
- -- brion
wikitech-l@lists.wikimedia.org