Improvements to the Category Walker

[haydenmuhl]

That's more than fair. This is why I put the caveat on my original post.

[pml]

That's fine, but it's similarly broken given my use case. If I am putting together a concert with my clarinetist friend, Auf dem Strom is going to be just as irrelevant a search result as Vocalise.

Yes, but if you’d taken the time to look, you would have found there is a unique category “for voice, clarinet, piano”. It’s easily found from the starting point I have given in each of my previous posts, and in fact I overlooked the easiest way to drill down from the generic to the specific — use the “subgenres” link instead of the “walk”. Want to guess how many works there are in it? (There’s apparently more than just Der Hirt auf dem Felsen. And it does not misleadingly include works like Auf der Strom or the Vocalise.)

As for tagging every single score as opposed to every single work — as Steltz suggested, it would easily occupy the team of librarians for perhaps two or three years without considering any new works or different editions of scores added in the meantime. It’s often impossible to determine correct tags without downloading the piece in question, and there’s no way I’m prepared to download everything (the entire 400 GB of the collection)!

In the mean time, the system does work and it actually works correctly and reliably if you invest a little time to examine how it classifies works. The complaint that “I am getting wrong results, the category walker is broken” sounds rather like someone who is disappointed to find that the volume of the encyclopaedia marked “S” doesn’t contain any entries starting with the letter “F”. Reflection might indicate there is another volume of the encyclopaedia sitting on the shelf that has been overlooked. I will agree the documentation for the end-user is probably lacking and a page or series of pages giving a tutorial on using the Category Walker system would help (or perhaps be of no interest at all; people just want to find the pieces they want). You perhaps weren’t around to see how badly broken the previous categorisation system was — there is no credible comparison with the new system.

Cheers, Philip

[haydenmuhl]

Yes, but if you’d taken the time to look, you would have found there is a unique category “for voice, clarinet, piano”.

I'm quite aware of these categories. IMO, they present problems of their own. For my use case, I'm interested in anything featuring voice and clarinet. Currently there are twenty four different categories that encompass these criteria, of which "For voice, clarinet, piano" only has three scores listed. That means there are at least twenty three other scores that feature voice and clarinet, but are not represented in that search. They go too far in the opposite direction, restricting what would be relevant results.

In the mean time, the system does work and it actually works correctly and reliably if you invest a little time to examine how it classifies works. The complaint that “I am getting wrong results, the category walker is broken” sounds rather like someone who is disappointed to find that the volume of the encyclopaedia marked “S” doesn’t contain any entries starting with the letter “F”.

Don't get me wrong. I think the category walker is a great tool. All I'm doing is suggesting an improvement. As to your analogy with the encyclopedia, I'm actually saying, "Why is the article for French Fries in the same volume as Solipsism?"

As for tagging every single score as opposed to every single work — as Steltz suggested, it would easily occupy the team of librarians for perhaps two or three years without considering any new works or different editions of scores added in the meantime.

Yes, I realize that my suggestion may not pass muster in light of a cost/benefit analysis. I still think it's a worthwhile discussion, though.

[imslp]

While I think this disjunct matching problem may be one of the few structural weaknesses of the CW system, I don't think it is entirely unfixable. Perhaps the easiest fix is to run a plain-text search over the MW:G list looking for categories that include both "Voice" and "Clarinet" (for example), and then show all the items in these categories, or even just to list the categories. Shouldn't be too hard. The question is how to design the interface (e.g. what types of searches are allowed).

[haydenmuhl]

Yes, I was thinking about something along those lines. If there were some way to convert the "For voice, clarinet, piano" tags into some other data structure that would match on a subset, it would fix the problem, even if they were still attached to the work page and not individual scores. That would also allow us to leverage the current categorization efforts, because the conversion from a "For X, Y, Z" tag to a {X, Y, Z} set could be automated.

[imslp]

I don't think there is a need to convert anything really: we can just do a very simple text search on each tag. The librarians have been very good at standardizing the category names, so that should work fine.

[imslp]

See: http://imslp.org/wiki/Special:GenreSear ... /clarinet/ (note the trailing slash)

[haydenmuhl]

Say, that's pretty spiffy. Did you throw that together just now?

What does the "Depth" column mean?

[ledbelly73]

I have a weird problem. I can't see any of the category walker functions on Safari, Firefox, or Chrome. I'm on a Mac Book Pro 2.33 GHz Intel Core2 Duo, OS X 10.6.7. A friend suggested holding Shift while clicking refresh, but it didn't help. Any suggestions?

[pml]

@ Haydenmuhl, I gather that’s exactly what Feldmahler just did (i.e. “throw it together”). I think depth refers to how far down within the category system a sub-category is nested: for example, you have large-scale top-level categories like “Vocal-Instrumental” – that’s a depth of 1; it includes categories like “for voices and solo instruments” which is depth 2; and then you drill down to “for voice, clarinet, violin, cello” or “for voice, clarinet, piano” which are depth 3.

@ Ledbelly, can you give us an example URL for a page you are viewing, where you then click on a link to access the Category Walker? An example might be, “I’m viewing the Mendelssohn page and I then go to click on the link for ‘Show works by type/instrumentation’.”

Cheers, Philip

[ledbelly73]

Thanks for your help. I'm on the "View Genres" page, where "walk" should appear on the left next to every entry, but it's not there. My ancient office G4 iMac shows it, so I think it's something particular to my computer. Much appreciated!

[pml]

Hi Ledbelly.

To populate those entries, the View Genres page uses a specific function call to the Category Walker called {{#imslpviewgenres}} which lists and formats all of the genre categories of a certain depth from one of the top level categories, and I’m not exactly sure how to debug what you’re seeing. Does the behaviour differ when you’re logged in as opposed to logged out? Have you had any success purging the page, e.g. Purge IMSLP:View_Genres?

Cheers, Philip

[ledbelly73]

Hi Philip,
Pardon my neophytedness on browser lingo. I have looked at the View Genres page logged in and logged out with no difference. I clicked the purge link you posted, but still nothing changed.

I'm also finding that the actions for the CW don't appear in other pages. I've tried clicking "Scores featuring the flute" and the CW link at the top of the page. In the Actions column there is nothing but blank boxes, where there was once the "restrict to" etc. functions.

Thanks for your help, I'm grateful.
Steve

[imslp]

Looks like it might be a system problem with monospaced font. Unfortunately I don't use a Mac, so I do not know how to fix it.

[ledbelly73]

Thanks for the hint on monospaced fonts, that led me to find a solution. If anyone asks something similar, it seems that Safari may not read Courier on IMSLP, but will read Courier New.
Thanks again.