Improving the accessibility of TNF knowledge - PLEASE READ!
#1
Posted 02 October 2004 - 09:38
Here, though, I'd like to ask for your help in something rather more specific.
How WE can help improve the performance of the Search Facility
Currently, searching for 'whatever' can produce vast quantities of results, right? And when one tries to narrow it down by adding a date, the software only recognises three digits and over. So, for example, if you search for 'Monza 78' your request gets vetoed. Bugger.
Still with me? Good.
So, despite there perhaps being a thread named (or including) 'Monza 78' you would have to sift through all the results for 'Monza' to find it. And there are a lot, which just wastes time...
Right - now for my suggestion!
From now on, as and when you come across a thread which includes a two digit date reference please send me the link and I will alter the thread title to include a four digit date instead. This will enable searches to pinpoint the thread we're after a lot more efficiently, thereby saving us from wasting our time and enduring unnecessary frustration.
Edit: this method can be implemented for marque type numbers too!
By doing this, we can help make the current technical set-up of the forum work better for us, instead of just sitting on our collective arse and whinging!
(Yes, I could go through the entire forum and do it myself. But I'm not going to! )
Advertisement
#2
Posted 02 October 2004 - 09:57
Originally posted by Twin Window
So, from now on, when you come across a thread which includes a two digit date reference please send me the link and I will alter the thread title to include a four digit date instead. This will enable searches to pinpoint the thread we're after a lot more efficiently, thereby saving us from wasting our time and enduring unnecessary frustration.
(Yes, I could go through the entire forum and do it myself. But I'm not going to! )
Great plan! But does this concern date references anywhere in any thread or date references in thread-titles only?
Mischa
#3
Posted 02 October 2004 - 10:04
Taking it one (entirely logical) step further to include dates within the content is by no means out of the question, it simply means more time spent for me!
Let's see how the title plan works out and I'll take it from there.
#4
Posted 02 October 2004 - 10:20
Or whatever specifics they're seeking to narrow it down, eg "Monza Peterson 1978"
Or are you saying it won't find, eg, "Lotus 20" for me?
#5
Posted 02 October 2004 - 10:30
It is now, David, since I merged the two main threads on that subject and re-titled it...Originally posted by David McKinney
I would have though it simpler for seekers to enter "Monza 1978"
The point is that it won't even perform a search for 'Monza 78' even if you know damn well that there's a thread of that name.
Geddit?!
Exactly, it won't - you'd have to sift through all the results for 'Lotus' unless you could identify another word or number of three digits or more in order to narrow the search down...Or are you saying it won't find, eg, "Lotus 20" for me?
#6
Posted 02 October 2004 - 19:12
And while we're at it - someone was asking for a spell checker....
#7
Posted 02 October 2004 - 21:27
The way I understand it, and Biranit of course will correct me if I am wrong, vBulletin already indexes everything it should index - it does not, however, index words with less than 3 letters, which is at least clever, as long as these 2-letter words are only marginal on English language. I would bet vBulletin does not index 'and', 'for', 'but', 'was', and a lot of other 3-letter words also, as they are just non-significant words, they just connect the ones that really matter. No Bulletin Board software can be written forecasting '23' is not just a number but perhaps the most succesful car ever built.
If the above statment is correct, merging Stu & Wolf's ideas is somewhat easy even though time consuming:
1. Any thread title with a 2-letter or 2-digit meaningful 'word' should be edited to include a 3 or 4 letters 'meaningful' something, like Stu already done with the Lotus T80 thread, and a new post, or the edition of any of the posts, should just include this new 3 or 4 letters 'something.
2. Any thread with any post including a 2-letter or 2-digit meaningful 'word' should be appended with a new post containg the 3 or 4 letters 'meaningful' something. And to to this no one needs admin power - you find it, you post on it. Dates will be easily driven this way, I guess.
3. This third step could be necessary or not - depending on the 3 or 4 letter 'somethings' that arise whilst steps 1 and 2 are done - but should be just a generic sticky helping searchers know, for instance, that 23 is M23, and also M16, and M19, and M24, and M25, and whatever. One can call this a 'cross reference serching tips' or just a 'puzzle yourself on it'.
I hope this can be of some help.
#8
Posted 02 October 2004 - 22:46
Originally posted by panzani
...vBulletin ...does not, however, index words with less than 3 letters...
OT maybe, but there seems also to have an upper limit in --depending on strings-- of 12 (eleven or or even ten).
You try Intercontinental and the only way to get it working is going down from the 16 original letters to something like *continental
The other way around (keeping the first part) leads you to go on discarding letters up to Intercont* -that´s ten if I count correctly.
#9
Posted 02 October 2004 - 23:08
#10
Posted 03 October 2004 - 04:55
#11
Posted 03 October 2004 - 05:37
Well yes, except that a 'Siffert' search - after dropping 'Jo' - would still yeild the results you wanted, as there weren't any other drivers of that name TTBOMK.Originally posted by Lec CRP1
I remember having this 'two character' search problem when searching for 'Jo Siffert' etc...
I have now merged the main threads concerning both the Lotus 72 and the Lotus 80 type numbers, which have been renamed by adding the 'T' prefix and thus creating a three-digit - and thereby 'searchable' - reference, along with 'Lotus'.
Therefore they can be now found by searching for Lotus T72 and Lotus T80. The old way of searching wouldn't have worked.
So please everybody, send me links for any threads which would a) benefit from a 2 digit to a 4 digit date change (and similar ones that could be merged with it), and b) links for threads where the type number needs to be made 3+ digits (and, again, links for any threads which could be merged with it).
Cheers
#12
Posted 03 October 2004 - 10:49
it seems not only, it is:
Originally posted by bira
We do not index words (strings) under 4 characters or over 12.
Truncation is IMHO no solution: when I search for Cunningham-Reid (the driver) it is not helpful to have to search for "Cunningham" (195 hits) or "Cunningham*" (210 hits)!
#13
Posted 03 October 2004 - 22:18
The way the search index works is that we define the minimum characters and the maximum characters for indexed words. We defined it as min 3 max 12 - and even that is pushing the envelop.
We got over 1.7 million posts on the Bulletin Board. To give you an idea: our entire BB database is 2.2 gigabytes. The "searchindex" table ALONE is 1.1gb - with 47,068,882 entries in it!
Anyone who is familiar with MySQL can tell you this is far from ideal, and over the winter we will find a solution for the BB, and the search.
It's very difficult right now to upgrade the BB. The easiest way would be to 'shut down' the existing BB and leave it as an archive, and start a new BB. But that would only solve the problem for future posts, not for the existing and past material.
Plus, I am not sure the newer vB version is any better.
The conclusion I have reached a while ago was that the BB should not have its own search engine. Rather, we need to install and run a separate search engine, with better capabilities (such as searching for exact phrases). Anyway, this is something we are working on and during the 2004/2005 off-season break should be addressing.
#14
Posted 04 October 2004 - 01:20
Originally posted by Twin Window
I have now merged the main threads concerning both the Lotus 72 and the Lotus 80 type numbers, which have been renamed by adding the 'T' prefix and thus creating a three-digit - and thereby 'searchable' - reference, along with 'Lotus'.
Thank You. From one who bleeds Hethel Green (or red,& white; black & gold etc etc) I appreciate your doing this. I've been adding the "T" prefix for months to the Lotus type number for just this reason.
#15
Posted 04 October 2004 - 07:31
"From now on, as and when you come across a thread which includes a two digit date reference please send me the link and I will alter the thread title to include a four digit date instead. This will enable searches to pinpoint the thread we're after a lot more efficiently, thereby saving us from wasting our time and enduring unnecessary frustration."
This certainly works for me.
Just one question really?
When are we going to adopt this method?
MCS
#16
Posted 04 October 2004 - 09:29
Great!Originally posted by Ruairidh
Thank You. From one who bleeds Hethel Green (or red,& white; black & gold etc etc) I appreciate your doing this. I've been adding the "T" prefix for months to the Lotus type number for just this reason.
Just as soon as I get sent some links!Originally posted by MCS
This certainly works for me.
Just one question really? When are we going to adopt this method?
And, of course, it makes sense for everyone to adopt this philosophy for all new threads & posts.
#17
Posted 04 October 2004 - 10:07
Originally posted by Twin Window
Great! Just as soon as I get sent some links!
And, of course, it makes sense for everyone to adopt this philosophy for all new threads & posts.
And, of course, it makes sense for everyone to adopt this philosophy for all new threads & posts.
Yes, this last bit certainly needs stressing!
MCS
#18
Posted 04 October 2004 - 15:18
I have been experimenting and I've discovered that a 'search entire posts' search includes the thread titles, which come up as the first post in the thread.Originally posted by Twin Window
Taking it one (entirely logical) step further to include dates within the content is by no means out of the question, it simply means more time spent for me!
This means that the data in the title provides the most important reference.
And yet only two people have provided me with links which would benefit from being merged, amended or both...
Come on, chaps!
#19
Posted 04 October 2004 - 17:43
I've just sent you a p.m.
MB
Advertisement
#20
Posted 04 October 2004 - 17:54
Thanks for getting involved...
#21
Posted 15 May 2006 - 14:21