Mailinglist Archive: opensuse-web (8 mails)

< Previous Next >
Re: [opensuse-web] Wiki search - proposal
On Mon, 02 Dec 2013 21:52:51 +0100
Christian Boltz <opensuse@xxxxxxxxx> wrote:

Hello,

I'm starting a new thread to make sure everybody who is interested
notices it ;-)

You all know that we get quite some complaints (and/or complain
ourself) about our wiki search because "it doesn't find the stuff".

This is mostly caused by the fact that by default the search only
lists pages in the main and Portal namespace.

I propose to change the wiki search and the config of Lucene (the
search engine used in the openSUSE wikis) to
- search all (relevant) namespaces
- use [Namespace-Boost] in lsearch-global.conf to boost the namespaces
that we currently search (main and portal)

The result will be that pages from main and portal namespace will
still be the top-rated search results, with the (currently hidden)
results from other namespaces below.


Here's a list of the namespaces we use, together with the boost value
I propose (1 means top rating, 0.0001 means it will be the last
search result, - means not to include this namespace in the default
search [1])

Boost Namespace number/name
1 0 Main
0.0005 1 Talk
0.005 2 User
0.001 3 User talk
0.6 4 Project ("openSUSE")
0.0005 5 Project talk
0.02 6 File
0.0005 7 File talk
0.005 8 MediaWiki
- 9 MediaWiki talk
0.0005 10 Template
- 11 Template talk
0.01 12 Help
0.0005 13 Help talk
0.02 14 Category
0.0005 15 Category talk
0.6 100 SDB
0.0005 101 SDB talk
1 102 Portal
0.0005 103 Portal talk
0.2 104 Archive
0.0005 105 Archive talk
0.2 106 HCL
0.0005 107 HCL talk
- 110 Book
- 111 Book talk
- 122 Property
- 123 Property talk
- 126 Form
- 127 Form talk
- 128 Concept
- 129 Concept talk
- 420 Layer
- 421 Layer talk

Just in case you wonder about the numbers - they are based on what
Wikipedia uses AFAIK:
(0, 1) (1, 0.0005) (2, 0.005) (3, 0.001) (4, 0.01), (6, 0.02),
(8, 0.005), (10, 0.0005), (12, 0.01), (14, 0.02)
and adjusted to our needs (see for example the difference in the
project namespace)

Sorted by importance

1 0 Main
1 102 Portal
0.6 4 Project ("openSUSE")
0.6 100 SDB
0.2 104 Archive - archive seems to be too high
0.2 106 HCL
0.02 6 File
0.02 14 Category - category should be .3
0.01 12 Help
0.005 2 User
0.005 8 MediaWiki - nothing for casual user
0.001 3 User talk
0.0005 1 Talk
0.0005 5 Project talk
0.0005 7 File talk
0.0005 10 Template
0.0005 13 Help talk
0.0005 15 Category talk
0.0005 101 SDB talk
0.0005 103 Portal talk
0.0005 105 Archive talk
0.0005 107 HCL talk




I'd like to know
- if you like this proposal in general,

Yes.

- if you agree with the list of namespaces to search by default,

Yes.

- if we want to include the talk pages in the default search and

No.
There is nothing there for casual user (reader, visitor).

- if you agree with the boost value for each namespace

Yes; see comments.

In the meantime, I'll fetch some cable ties to tie up Henne - I'm
sure he won't like the idea of including all those namespaces in the
default search ;-))

Current distro release presentation should go to
http://opensuse.org/distro/
and wiki should become what it is - a wiki.

Needless to say that we might need to finetune the boost values of
some namespaces later (needs some testing with the real wiki
content), but in general I think the above values will work much
better than what we have now ;-)

I would just apply those to github/wiki and check how that works.


Regards,

Christian Boltz

[1] Nevertheless, we should set those namespaces to a boost value of
0.0001 in case someone includes them in the search



--
Regards, Rajko.
--
To unsubscribe, e-mail: opensuse-web+unsubscribe@xxxxxxxxxxxx
To contact the owner, e-mail: opensuse-web+owner@xxxxxxxxxxxx

< Previous Next >
List Navigation
References