Open main menu

User talk:Jc86035

About this board

Previous discussion was archived at User talk:Jc86035/Archive 1 on 2016-10-01.

Thinking about a new property: "recorded versions"

10
Moebeus (talkcontribs)

Something similar to tracklist, but for compositions of course.

  • It's a concept/term that is in wide use on (English) wikipedia.
  • It would function as a reverse property of Property:P2550 "recording or performance of"
  • it would allow linking a musical work straight to a music track without having to go through our current, slightly ad-hoc method of Performer>statement is subject of.
  • For the sake of readability properties like "performer" and "publication date" could be optional

Thoughts?

Jc86035 (talkcontribs)

Maybe not. I think it's unlikely that any more inverse properties will be approved, regardless of topic.

Moebeus (talkcontribs)

Sorry if I'm harping on about this, but could Property:P1811 be used as an argument for "list of recordings" maybe?

Jc86035 (talkcontribs)
Moebeus (talkcontribs)

huh. Maybe I'm reading that wrong, but I can't see any reason given why inverse properties are bad practise, other than the commenter not liking it?

I realize of course that inverse properties aren't technically necessary, but that ignores the fact that people will keep adding properties they can't actually read/see on the item itself. So from a UX perspective I think they absolutely serve a purpose.


P.S. "There's a piece of user-written javascript you can use to automatically see properties on an item from the reverse direction" - do you know what userscript he refers to here, it's not the "What links here"-link is it?

Jc86035 (talkcontribs)

I think the user script is Wikidata:Tools/Enhance user interface#derived statements. It appears to be somewhat useful, although it doesn't show qualifiers or references.

I think the necessity of the inverse property primarily depends on whether the MusicBrainz review/import is completed before or after inverse property querying is enabled in MediaWiki. (Given that m:Community Wishlist Survey 2019/Wikidata/Automatically add inverse properties was doing surprisingly well until it was pointed out that the specific method of the proposal (adding all the inverses) was much more problematic than just allowing inverse database queries, I would expect a similar proposal based on querying to do well in the November 2019 survey, so I'm guessing that the implementation would be between about 15 and 24 months from now.)

Anyway, regarding ArthurPSmith's comments, DeltaBot recently added about 1,100 child astronomical body (P398) inverse statements to Sun (Q525). If I were to propose the property then I would emphasize the argument that it's stupid to allow some functional inverses but not others (though the utility of an inverse recording property would probably be higher than that of an inverse modified version property).

Jc86035 (talkcontribs)

Finally, regarding the quality of MusicBrainz, I'm currently in the process of archiving a very large amount of Spotify albums to the Internet Archive, and extracting the ISRCs from tatsumo.pythonanywhere.com. Having a complete static database of several million ISRCs, connected to Spotify identifiers, it would probably become at least a little easier to fill out the gaps in MusicBrainz's data.

Moebeus (talkcontribs)

Holy Moses, that's awesome! I don't know how you find the time to do all this stuff I'm well impressed πŸ‘πŸ‘πŸ‘

Jc86035 (talkcontribs)

Me neither, to be honest. I said I was going to take a break soon. (Fortunately for me the archival work is largely automatic.)

Moebeus (talkcontribs)

I selfishly hope your break won't last very long, but you certainly deserve it ;)

Reply to "Thinking about a new property: "recorded versions""

{{P|435}} conflicts with {{P|436}}

12
Mineo (talkcontribs)
Jc86035 (talkcontribs)

My decision was intentional, yes. The 2015 omission was also a presumably unilateral decision, by Nikki, and this was several years before Moebeus (and I) began separating singles into items which would be one-to-one matches with MusicBrainz entities. All Wikidata users, including unregistered users, can add or remove property constraints.

While Wikidata may not in practice distinguish songs from singles (except when noting covers of a song), it is probably beneficial nevertheless to have different Wikidata items for the distinct entities. (See also the deletion discussion for B-side (DEPRECATED) (P1432).) Unfortunately, the data is terribly inconsistent at the moment, because there are more than 80,000 items for singles, and so far all of the several hundred separations have been done manually. In the meantime, data reusers will have to deal with it or use other databases.

Moebeus (talkcontribs)

@Mineo There is another issue with your bot: It updates WD not only based on WD links on MusicBrainz, but also based on wikipedia-links. This breaks a lot of stuff, and I think your updates should be limited to WD-item<>MB item. The reason for this is that Wikipedia articles will often cover several subjects (song, cover versions, releases) and MusicBrainz editors will link wikipedia articles to display relevant text on their items, which is fine. The problem comes when these links are imported back into Wikidata, like what happened here: Q7428899 - in this example two versions of the same song are correctly tagged with WD items, but the wikipedia article (which covers both versions) causes problems. Could you have a look at your script?

Jc86035 (talkcontribs)

This is a known issue. I believe the answer is that because Wikidata has functional constraints and querying infrastructure and MusicBrainz does not, the error-fixing should occur on our end and the same fixes should then occur on MusicBrainz.

Mineo (talkcontribs)
Moebeus (talkcontribs)

Would it be possible to check if an MB Release/MB Release group ID is already present on the item before adding a MB Work ID? Personally I think the biggest problem is items with more than one/conflicting MB IDs, but that's just my opinion.

Jc86035 (talkcontribs)

If you go to a random category like w:en:Category:1988 singles and check the Wikidata items, you'll find that basically all the data was imported from the English Wikipedia infoboxes in 2012 and 2013. I don't think a lot of those are linked to MusicBrainz at all.

Jc86035 (talkcontribs)

I think adding them will still be beneficial, since (depending on your point of view) the item represents the recording and composition as well as the single, since all of the English Wikipedia articles are supposed to have the song as the primary topic unless the single has more than one track (I think?). (The data is already polluted, in any case, so MineoBot won't be breaking anything that wasn't already broken.) I don't think there's anything official, and there are few active contributors in this area, but for now I think adding it should be okay until someone comes around to clean up all of the items.

Mineo (talkcontribs)
Jc86035 (talkcontribs)

The template parameter was added by Laddo in 2014, which was shortly after the Dexbot import in 2013. At the time this would have been accurate, and at the time property documentation was always stored in the template on the property talk pages instead of being indicated in the Property namespace. I didn't even notice that text until you mentioned it.

I've removed the template parameters on the talk page, since they are redundant to and/or contradict the information stated in the property entity itself. Everything here was done unilaterally (I think), and a revert and a discussion should be enough to establish consensus.

Mineo (talkcontribs)
Jc86035 (talkcontribs)

As always. I think Wikidata right now is a bit like an emptier 2006 Wikipedia (no references, everything is a bit messy). I wasn't there in 2006, but Wikidata's about the same age as Wikipedia was in 2006.

I think this comment I made yesterday might be relevant. Essentially, there are 80,000 to 500,000 items to fix (across music and books), and I don't really know how that would be done because I haven't learned how to run a Wikidata bot and don't know if it would be possible for me to do it.

Reply to "{{P|435}} conflicts with {{P|436}}"
Moebeus (talkcontribs)

Just saw this linked on the project chat: Talk:Q14211, never seen that done before. Wouldn't that be a great way to list cover versions without cluttering up the Item with the "performer/statement is subject of" model? Haven't really thought it through or anything, do you see any immediate drawbacks implementing something like this?

Jc86035 (talkcontribs)

In the short term, I think the template itself could be beneficial, although bot code had to be written for it and I don't know if anyone will want to do that. Item talk pages are still mostly a barren wasteland.

On the other hand, removing the statements would probably have some knock-on effects. Since it would currently prevent infoboxes from using the related data, it does depend on whether it's likely that inverse statement access (phab:T209559) will be implemented before Wikidata infoboxes gain steam. (It's likely that inverse statement access could get top 10 in the community wishlist survey (see task comments), which would effectively put its deployment around late 2020; and I don't see Wikidata infoboxes gaining steam at all in the big (i.e. relevant) Wikipedias until there's a consistent level of adequate referencing, which would probably take several years even with supporting efforts like GlobalFactSync.) It's impossible for us to know one way or the other.

Anyway. You already have User:Pasleim/derivedstatements.js installed, so you can already easily check whether there are any inverse statements and go from there. I probably wouldn't rely on waiting for any of the aforementioned things to be resolved.

Reply to "Have you seen this?"

musical work / musical composition

4
Moebeus (talkcontribs)

regarding https://www.wikidata.org/w/index.php?title=Q1886329&oldid=prev&diff=885609086

"I'm not totally sure how the ontology works here, but using musical work (Q2188189) breaks some identifier constraints, and it's sort of strange to imply that somehow this isn't a composition"

What constraints are broken? I've noticed there has been a bit of activity with edits and reverts on musical work and musical composition. In theory, they should be almost synonymous but that might have changed recently. Any item with an ISWC-identifier is, per definition, a musical work?

This question comes up a fait bit (see User talk:Andrew Gray#"musical work" / "song" ) - maybe we could define it a little better over at Wikidata:WikiProject Music to avoid confusion.

Jc86035 (talkcontribs)
Jc86035 (talkcontribs)

There's no English Wikipedia article for musical work (Q2188189). The German Wikipedia article states that "Musikalisches Werk" encompasses both compositions and collections of such compositions (such as albums). The second paragraph essentially states that any work containing music is also considered a musical work, including those that also include sound or imagery (e.g. an album).

Moebeus (talkcontribs)

I'll look into that constraint later, that definitely shouldn't be triggered.

I don't mind using musical composition (rather than work) or in any way think it's wrong, my main thing is that just using "song" is not enough as long as song and single are so intertwined.

Reply to "musical work / musical composition"
NMaia (talkcontribs)

Hey! This is sort of off-topic from the discussion so I'm sending this privately. I think there are things to be improved in the metro as can be seen in this query. Your other one seems alright.

Jc86035 (talkcontribs)

For the first query, those are indeed the only metro stations in Tsuen Wan District (Q878514). All of the adjacent stations are in other districts. I tried changing the query to use MTR station code (P1377) (which is definitely used on all MTR stations) and that worked a bit better, although I haven't read through the whole query so I don't really know how it works.

Reply to "Queries as QA"
Moebeus (talkcontribs)

Do you have any ideas how to best model mashups?

Have a look at this if you have time, I can't get seem to get it right: Q7459782


Jc86035 (talkcontribs)
Moebeus (talkcontribs)

Yeah, I've leaning on statement is subject of (P805) way too much, almost like a catch-all, after someone told me "instance of" shouldn't be used as a qualifier. I've come to over-use/misuse "object/subject has role" as well.

Jc86035 (talkcontribs)

Perhaps the qualifier isn't necessary? The item already clearly indicates that it's based on more than two originals and links to mashup (Q48445) and remix album (Q963099) in other statements.

Reply to "how to model mashups"
Moebeus (talkcontribs)

re your flexi disc edit: is it em dash (β€”) or horizontal bar (―) you're using? I didn't know about this convention, will stick to that going forward, thanks.

Jc86035 (talkcontribs)

En dash (–). I'm not sure how common it is in languages other than English, though, so it might not be appropriate to use it in all of the labels.

Reply to "em dash"
Moebeus (talkcontribs)
Jc86035 (talkcontribs)

The "precedent" is more like the usual "we imported this and no one really cared that it wasn't technically correct so we kept going", isn't it? A single is literally not "part of" an album.

The notion of "single from an album" is fairly complicated, I think, and it doesn't seem very consistent to me, even for the more popular albums. For example, w:en:Dua Lipa (album) lists eight of the eleven singles containing tracks from the album as "singles". (The three that aren't listed are only included on the "complete" version, which was released about a year after the original album.) However, the oldest of those singles was released almost two years before the album itself was. On the other hand, w:en:Red Pill Blues lists only the latter three of the five singles containing tracks from the album; the second single was released nine months before the album was, and the third single was released two months before.

The rules for such a distinction might also vary from region to region; and if there were a complete set of data for an artist, it could be possible to query for singles associated with an album depending on how intricate the given rules are.

If there are actual definitions I think sticking more abstract concepts (e.g. "release for the purpose of album promotion" with an of (P642) qualifier – using of (P642) on the main instance of (P31) statement is probably too vague) into instance of (P31) could work, although this is really not anywhere near the top of Wikidata's music-related priorities.

Moebeus (talkcontribs)

"The "precedent" is more like the usual(...)" - I can't argue with that. I actually toyed with the idea of Album=>"has list"=>"list of singles taken from the album"=>list of singles, but abandoned it since it seemed like overkill for anything like super smash hit albums like Michael Jackson's Thriller where almost every song got released as a single.

Do you suggest simply removing these links as I find them maybe? I agree that this shouldn't be a priority, but it is an area where people seem to enjoy contributing 😐

Jc86035 (talkcontribs)

I think it's probably better to remove them for now (at least in their current form; and I would have thought the part of (P361) was supposed to refer to the track and not the single), but I'm not sure. On the English Wikipedia there are usually mild edit wars over what tracks were actually released as singles (since e.g. there's no technical difference between a "digital single" and a "digital promotional single") and which ones count as "from the album".

Reply to "Andy Gibb"
Moebeus (talkcontribs)

Just looking for some guidance here: shouldn't Property:P1725 only be present on recordings, and not on the compositions themselves?

I thought Property:P1558 and Property:P3440 where for compositions while BPM was typically derived once a recording is made?

Jc86035 (talkcontribs)
Moebeus (talkcontribs)

I can't read music unfortunately - was that an example of both being present? I thought BPM was more of a modern term adapted to suit electronic music, boy does it look like I was wrong, hehe.

I guess what I'm after is a way to separate the time/tempo the composer envisioned, from recorded versions that might slow an original down or speed it up, like the Smurfs Eastern European pop stars are wont to do. Those guys often won't bother with registering an arrangement.

Jc86035 (talkcontribs)

Yes, the score's tempo marking is both in words (Allegro moderato – trΓ¨s doux) and in BPM (120 quarter notes per minute). The piece is from 1903, although it seems BPM is a consequence of the metronome (1810s).

For recordings without sheet music I don't think it's worth adding the tempo, since it's practically unverifiable (from Wikipedia's standpoint) and any consequential addition of that data would probably have to come from a massive algorithmic MusicBrainz import. I would probably just add the length of the recording (to the recording) and the number of bars in the original sheet music (to the composition), since then the BPM could be approximated with bars Γ— time signature Γ· duration in minutes.

Jc86035 (talkcontribs)

(To be honest, I'm not really sure if any import could be made since all of the relevant databases are probably copyrighted. I don't have much experience importing data, but it's probably not something that anyone will really want to import in the near term, anyway.)

Moebeus (talkcontribs)

I see there's a few databases specializing in BPM: https://tunebat.com/, https://songbpm.com/, etc. If I'm not mistaken it's also part of the Spotify API, if they haven't closed it down by now. Well never mind, I guess I was just inspired by the tremendous work you guys have done on Q166904 Someone Like You. Seriously, that blows my mind right there, top notch πŸ‘πŸ‘πŸ‘

I'll get back to what I'm actually any good at - tracking down the originals!

Reply to "P1725 beats per minute"
Moebeus (talkcontribs)
Jc86035 (talkcontribs)

Sorry, I didn't notice you'd posted this. I haven't used some of those properties much, really, but they do all seem to be distinct (even if difficult to choose) to me

Moebeus (talkcontribs)

Dude, I'm trying to go through my watchlist from when I was gone and I can see you've done a massive job reverting bad edits this last weeks. Big thanksπŸ‘

Reply to "so many choices, so little time"
Return to the user page of "Jc86035".