Search problem

[Amphissa]

The search function seems to be misbehaving.

A recent item in Downloads, Sydney Grew said "let us have at least one item from Bohuslav Martinů." Which seemed odd, since I had posted a Martinu item just last month for download.

So I ran a search for Martinu, both in the Downloads section and the site at large. Sydney's post was located, but mine was not. I checked, it's still there on page 6 in the Czech folder.

So, why is the Search function finding his and not mine?

Another example from the Czech folder is Novak. If you search site-wide or Downloads as a whole, you don't pick up the May Symphony, Notturna and other items by him.

I've run across this problem before, but not reported it and don't remember the instances. But it does cause concern. I often search to determine whether something is available, but also to see if something I want to upload has already been posted.

[Dundonnell]

I appreciate the nature of your frustration with the Search facility

With reference to your last sentence...there are at least now indexes of all the British, Irish, American, French and Czech works which have been uploaded for members of the site to allow you to see whethersomething is available or has already been uploaded. With the exception of the Czech catalogue these are all now stickys and perhaps the Czech index(Sydney Grew's post in Czech Downloads Discussion of 13 April 2012) should be added as another sticky

[fr8nks]

I have made at least 2 posts and sent 1 private message suggesting that the search facility is flawed. I had instances to support my findings but now don't remember them. I think you all do a great job and I wouldn't want to be responsible for the upkeep of the site but I agree with Amphissa that the system is not perfect. I could, upon request, find examples where the search facility is incomplete.

[fr8nks]

An example which just came to mind is the download search for Jan Hanus. Only Symphony No.5 appears in the search facility but I believe all six symphonies have been uploaded plus a ballet suite.

[TerraEpon]

From what I can tell, it only shows the last post made with the word in each thread. Pretty dumb, yeah.

[Mark Thomas]

Well, the the software is off-the-shelf open source software and the search function is what comes built in. It's main idiosyncrasy, as I have said many times before, is that it is hierarchical and so, if you type in the search box when you're in a board, it only shows results from that board. That's why I always recommended searching from the home page or the Search page itself. All that said, I appreciate that the inconsistencies which you've highlighted aren't explained by that so I'll see if there are any diagnostics available to test the accuracy of it's results and I'll poke around the support forums in the hope that someone out there has had a similar problem and can fix it.

[hemmesjo]

I don't know if this has anything to do with it but each of the names has a non English letter in the original spelling.  Martinu (Martinů), Novak (Novák) and Hanus (Hanuš).

Dan

[JimL]

Oh, great!  And unless you have some sort of guide to what's in your computer's character map you could spend hours trying to find the right key to hit! 

[jerfilm]

Betcha Dan's got it.......

Jerry

[hemmesjo]

How about a list of names with the original spelling that could then be copied and pasted as needed?

Dan

[Amphissa]

That doesn't seem to matter for the Search facility. Type Martinu in the Search box, you get SG's post, which is spelled Martinů, but you don't get mine, which is spelled Martinu.

[hemmesjo]

I just did two searches from the opening page.  One was Martinů and the other Martinu.  I received entirely different responses.

Dan

[Mark Thomas]

I've run the diagnostics and they show that all is in order. That said, the default setting for the search index is aimed at optimising the speed of search by limiting the size of the index (by taking out common words like "the", "and" etc.) and that can apparently prejudice the accuracy. So I'm currently re-indexing to produce a much larger index in the hope that it will improve the accuracy without slowing down searches too much. As to accented characters, I can find no references to the way Search treats them either within the software settings of the search function, or in the online support pages. I'll keep looking but the default in this situation seems to be that á is treated as a and vice versa. Quite how it would cope with ß I don't know! Anyway, this is work in progress. I'll post here when and if I find out more....

[Mark Thomas]

I've now replaced the default search index with a much more comprehensive one, which should make searches more accurate and doesn't seem to slow down the search function. Unfortunately, it still doesn't see Amphissa's Martinu post when you use "Martinu" as the search term, although it does when you use "Julietta". Very strange. I supect that accented characters aren't treated as unaccented ones and vice versa, but I have asked the question in the software support forum and we'll see what response we get. Finally, I have added a modification to the board which adds a drop down dialog next to the search box, so you can choose how much of UC your search will cover. This should solve the issue of only being able to search the whole board from the home page. More later, if there are any developments.

[MikeW]

What wildcards are supported? I tried ? and % but got variable results (pun not intended).

Also I keep forgetting that the Prev/Next at the bottom of each page don't move between pages but threads. Rather peculiar.