Clouds
|
View:
New views
15 Messages
—
Rating Filter:
Alert me
|
 |
Clouds
by Simon Reinhardt
::
Rate this Message:
Hi,
in chat, nikki talked about an idea she had: <nikki_> I'd like a way of associating albums with a *thing* <nikki_> like... <nikki_> a group which contains all the ghost in the shell CDs, soundtracks, singles, etc. <nikki_> but since it's not the artist, and many artists created music for the films/TV series, they're all over the place <nikki_> (it would have the advantage of people who don't give a toss who the composer was having a pseudoartist to use too) That gave me to think and when I was editing around various progressive rock artists who have a lot of side projects I realized the following: * We build discography pages for single artist names only. And we add ARs to relate them to other artists and works of them. You have to click through the ARs to see the discographies of band members, side projects, to see songs they featured on, albums they produced, songs they remixed and so on. This is all very much unordered of course and you have to click on "View all relationships" and look through dozens of identical songs. While ArtistPageRedesign aims to better this and NextGenerationSchema helps with those problems too by grouping tracks, it will still be hard to give a general overview of related works and artists. Better would be a page listing combined discographies. But what to combine though? * Similar problem with jazz artists. Artists perform in a lot of lose bands, play sessions together and so on. So you have a lot of artist entities in the db for one jazz professional. Like Duke Ellington or Count Basie. * There is nothing relating VA releases at the moment. The AdvancedEntity page proposed new entity types for compilation series and for labels. With labels you can find albums by catalogue numbers and series entities would help to keep an overview to huge series like Café del Mar and the like. The following concept is more open though. Like Don would say: it's less hard structure of the data (by db means) but more soft structure (through guidelines). Basically the idea is that users can create so called "clouds" to combine artists, albums and tracks. This would be - based on the wiki idea - a very open and chaotic system with very soft and accessible linking methods, not hard links as AR does it. Though the linking would be very flexible through some nice features. An example: I want to create a discography for the drummer Mike Portnoy. So I go and create a new cloud object and tell (with a special syntax) it to include every album that: - is listed under Mike Portnoy, - is listed under one of the bands or collaborations he was part of, - he otherwise played on. I give the cloud a title and others can find it when searching for him (separated from the normal search results). Other linking methods would be: albums he produced, tribute albums he has aso. Such a list could also be done for tracks. The special syntax would be an 'intelligent linking method'. But you could also just add single albums. For compilation series I would also create a cloud to have a container for all albums. This could be just adding single albums to the list or by creating an intelligent search filter. For example I want to list all the albums of the Real Ibiza series. So I add a search filter "real ibiza". That way I'd have every album that has this in its title - better than manual adding them because if new albums are added to the db they appear on this list and you can check if there are duplicates. This album list also contains albums from other series though, like "Chillout Dreams 2: The Real Ibiza Grooves". You would filter that out with a blacklist. A rough idea of such a syntax: include:artist:<id>:albums include:artist:<id>:memberof:albums include:artist:<id>:collaborations:albums include:artist:<id>:production:albums include:album:<id> include:albums:*Real Ibiza* exclude:albums:Chillout Dreams* It would then of course need an user interface to create that. Note that this is just an abstract idea that came to my mind. I'm not proposing it for MB but just ask for a little brain storming. :) Perhaps I can code a simple version of it if I ever come to write a Java API for the new webservice. Simon (Shepard) _______________________________________________ Musicbrainz-experts mailing list Musicbrainz-experts@... http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-experts |
|
|
|
|
|
Re: Clouds
by Simon Reinhardt
::
Rate this Message:
Matthew Exon wrote:
> Simon Reinhardt wrote: > >> Note that this is just an abstract idea that came to my mind. I'm not >> proposing it for MB but just ask for a little brain storming. :) >> Perhaps I can code a simple version of it if I ever come to write a >> Java API for the new webservice. > > I think of MB not so much as a website as a resource. Go create your > Mike Portnay web page on your own server, but suck in the Discography > information via MB web services. MB concentrates on the things that fit > into its schema, which keeps it tight and reliable, while the XML web > services format provides whatever flexibility you need. Jup, that was my idea: first do something external. Luckily with the new webservice this will be possible. > The syntax you suggest reminds me of SPARQL. I don't know if SPARQL is > common knowledge on experts already. It looks like you want to do a > query like this partially pseudocoded version: Well it's not really a suggested syntax, more that you can see what I have in mind. I will take a look at SPARQL, sounds interesting. Simon (Shepard) _______________________________________________ Musicbrainz-experts mailing list Musicbrainz-experts@... http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-experts |
|
|
Re: Clouds
by DonRedman
::
Rate this Message:
On Tue, 28 Feb 2006 00:55:32 +0100, Simon Reinhardt wrote:
> Note that this is just an abstract idea that came to my mind. I'm not > proposing it for MB but just ask for a little brain storming. :) > Perhaps I can code a simple version of it if I ever come to write a Java > API for the new webservice. I like it. Basically these are tags, just like last.fm uses them. You can tag any MB Object with a string. That's all there is to it. Your syntax is just an aid to tag many opbects that have certain properties in one go. The interesting question with tags is, under which conditions can people add them and change them. AFAIU delicious, flickr, and folksonomy in general it is crucial to bind the tag to the user who _uses_ it. A tag is not simply "this element is tagged as 'xxx'" but "user 'xy' tagged this element as 'xxx'". this way you can filter tag by just using - your tags - everybody's tags - everybodys tags except from people you consider "spammers" And this makes a moderation system obsolete. DonRedman -- Words that are written in CamelCase refer to WikiPages: Visit http://wiki.musicbrainz.org/ the best MusicBrainz documentation around! :-) _______________________________________________ Musicbrainz-experts mailing list Musicbrainz-experts@... http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-experts |
|
|
Re: Clouds
by Steve Wyles
::
Rate this Message:
On Tue, 28 Feb 2006, Don Redman wrote:
> AFAIU delicious, flickr, and folksonomy in general it is crucial to bind the > tag to the user who _uses_ it. A tag is not simply "this element is tagged as > 'xxx'" but "user 'xy' tagged this element as 'xxx'". this way you can filter > tag by just using > - your tags > - everybody's tags > - everybodys tags except from people you consider "spammers" > > And this makes a moderation system obsolete. > Erm, no. It doesn't replace the moderation system. All it provides is various views upon the data, one of which *might* provide authoritive information. Look at the mess freedb is in, because different people want their album to look a particular way. Steve (inhouseuk) _______________________________________________ Musicbrainz-experts mailing list Musicbrainz-experts@... http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-experts |
|
|
|
|
|
Re: Clouds
by Simon Reinhardt
::
Rate this Message:
Matthew Exon wrote:
> [possibly redundant given Steve's response; but I wrote more, and used > bigger words :-P] > > Don Redman wrote: >> On Tue, 28 Feb 2006 00:55:32 +0100, Simon Reinhardt wrote: >> >>> Note that this is just an abstract idea that came to my mind. I'm >>> not proposing it for MB but just ask for a little brain storming. :) >>> Perhaps I can code a simple version of it if I ever come to write a >>> Java API for the new webservice. >> >> I like it. Basically these are tags, just like last.fm uses them. You >> can tag any MB Object with a string. That's all there is to it. Your >> syntax is just an aid to tag many opbects that have certain >> properties in one go. > > Yes, that's a better way of looking at it. My mind is on semantic > networks right now, so I see everything as "subject-predicate-object". > But really the only predicate Simon suggested was "is something to do > with", so that boils down to tags really. Hmm. I feel a bit misunderstood. :) The semantic relations are already there (AR, normal artist-album-track relations). My suggestion uses them to build subjective lists that one can create to provide an alternate view on the data. The purpose of this is either enhancing the search, so one can more easily find connected data, or to create an overview over complex spreaded discographies. This is not really connected to tags. With 'clouds' I don't mean 'tag clouds'. You create a list of artists/albums/tracks by either manually adding single items or by letting a smart filter automatically include items (this is of course not fixed then, it does a lookup everytime you access the list) and then give the object you created a certain name. I don't think that's the same as taking items and adding tags to them. > It's exactly this function that makes MusicBrainz the most advanced > community database I've ever seen. Because it's a full-on voting > system, it's way more advanced than anything Wikipedia has, and thus > avoids the "last edit wins" problems. Survival of the fittest promises > to make it even better. I said this on the summit but I'll repeat it here for everyone. :) I don't think SurvivalOfTheFittest still fits on the complexity MB has reached. It is an out of date proposal in my eyes. We talked a lot about new moderation systems on the summit and came to the conclusion that we can divide between two kinds of edits: Structural edits (add, delete, move, merge and the like) and textual edits (any kind of changes of text data fields). Given that we pondered wether a future system could keep the votings for structural edits and use a wiki editing system for the text type edits. But I think even for textual edits only the SurvivalOfTheFittest paradigm fails. I saw an edit today where it would have been nice to have it: http://musicbrainz.org/mod/search/results.html?object_type=album&orderby=desc&object_id=22702 And then I saw one where it would end in absolute chaos: http://musicbrainz.org/showmod.html?modid=4347859 - imagine people voting 2:1 for the one title on this track and other people voting 0:2 for the alternative title on the other tracks that are on other albums. Then we have chaotic data. Yeah of course this should be self-regulating over time. But will it in practice? IMHO reasonable automoderators deciding for one alternative (after discussions or not) and then making the data consistent through all albums are much better. I hope we can redefine the consistency philosophy after introducing the NextGenerationSchema a little though... Simon (Shepard) _______________________________________________ Musicbrainz-experts mailing list Musicbrainz-experts@... http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-experts |
|
|
Re: Clouds
by DonRedman
::
Rate this Message:
On Tue, 28 Feb 2006 16:40:38 +0100, Steve Wyles wrote:
>> And this makes a moderation system obsolete. >> > > Erm, no. It doesn't replace the moderation system. All it provides is > various views upon the data, one of which *might* provide authoritive > information. Look at the mess freedb is in, because different people > want their album to look a particular way. Yes it does make moderations obsolete if (and only if) you view clouds as *tags*. Tags are not objective facts. They are subjective labels which get their validity from the fact that someone uses them. What last,fm and del.icio.us do, is to statistically analyse the subjective labels of thousands of users. The 'mean' of this statistical analysis is not 'objective' but gets it right most of the times, because it reflects the semantic concepts of the majority. Tags are always something vague. I realise that Simon's original idea is not vague in this sense. He wants to construct a very specific clound that he calls "Mike Portnoy". Now there are two ways to achieve this: a) Have a voting system and let people debate and vote which elements belong to "Mike Portnoy" and which do not. At the end you will end up with a cloud based on the *oppinion* of the majority. b) Just *use* tags. Let people tag all those things that belong to "Mike Portnoy", according to their *usage* (not *oppinion*, user A might just tag those albums that she owns). If you make a statistical aggregate of all these tags, you will end up with a could based on the *usage* of the majority. My point is that while the second option cannot claim to be objective, it is as useful as the first one. In this context common useage is just as good as a criterion for validity as democratic vote. And it removes the need for voting, delaying changes to the db and all that stuff. DonRedman -- Words that are written in CamelCase refer to WikiPages: Visit http://wiki.musicbrainz.org/ the best MusicBrainz documentation around! :-) _______________________________________________ Musicbrainz-experts mailing list Musicbrainz-experts@... http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-experts |
|
|
Re: Clouds
by Steve Wyles
::
Rate this Message:
On Tue, 28 Feb 2006, Don Redman wrote:
> On Tue, 28 Feb 2006 16:40:38 +0100, Steve Wyles wrote: > >>> And this makes a moderation system obsolete. >>> >> >> Erm, no. It doesn't replace the moderation system. All it provides is >> various views upon the data, one of which *might* provide authoritive >> information. Look at the mess freedb is in, because different people want >> their album to look a particular way. > > Yes it does make moderations obsolete if (and only if) you view clouds as > *tags*. > > Tags are not objective facts. They are subjective labels which get their > validity from the fact that someone uses them. What last,fm and del.icio.us > do, is to statistically analyse the subjective labels of thousands of users. > The 'mean' of this statistical analysis is not 'objective' but gets it right > most of the times, because it reflects the semantic concepts of the majority. But, musicbrainz is about recording the factual information for the releases, subjectiveness doesn't come into it. Either something is correct or it isn't. If something is factually correct in the database, I want it displayed as that 100% of the time, not adjusted by the result of some statistical element. With this idea of tags, the incorrect data in the database would always be incorrect because nobody is using it. The moderation system currently allows somebody to correct that data, if they wish, whether it is used or not. Then in the future if somebody wants to use it, it will be correct. Remember, there are other people using the data (data feed customers) where their users don't have direct interaction with the musicbrainz site. Please think of the bigger picture regarding the data, not just the use of it via the musicbrainz software or website. Steve (inhouseuk) _______________________________________________ Musicbrainz-experts mailing list Musicbrainz-experts@... http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-experts |
|
|
Re: Clouds
by DonRedman
::
Rate this Message:
On Tue, 28 Feb 2006 18:13:29 +0100, Steve Wyles wrote:
> Please think of the bigger picture regarding the data, not just the > use of it via the musicbrainz software or website. Oh, I am. The picture I am thinking in includes all users of MB data, what they do with it, and how their actions can be turned into value for MB and other users. > On Tue, 28 Feb 2006, Don Redman wrote: > >> Tags are not objective facts. They are subjective labels which get >> their validity from the fact that someone uses them. What last,fm and >> del.icio.us do, is to statistically analyse the subjective labels of >> thousands of users. The 'mean' of this statistical analysis is not >> 'objective' but gets it right most of the times, because it reflects >> the semantic concepts of the majority. > > But, musicbrainz is about recording the factual information for the > releases, subjectiveness doesn't come into it. Either something is > correct or it isn't. If something is factually correct in the database, > I want it displayed as that 100% of the time, not adjusted by the result > of some statistical element. Until now MB has only stored *itemized* *objective* data. IMO Simon's proposal crosses an important line, one which we know that we will eventually have to cross. On the other side of the line you find *aggregated* *subjective* data. What I am saying is that the logic of itemized objective data does not apply to this realm. If we force our old logic onto this new type of data, then we loose at least half of the power of that data. Yes, "'... rocken tun die andern!' is an album by 'gutfürnkeller'" is either correct or wrong. This information is objective. It state a fact about two items. But whether gutfürnkeller's music is *called* 'Funk', or 'Jazz', or 'nuJazz', or 'Blues' is not objective (people's oppinions or personal classifications differ), nor does it state facts about an item. Rather it provides *contextual information* for some items. The power of what people call the "Web 2.0" (and they call it like that because they need a name for a phenomenon that they do not understand) is that these contextualizations are aggregated and this aggregate is made publicly available. There is nothing objective about this aggregate (it is an aggregate of many people's subjective contextualizations) and it does not state a fact about a single item (it gives some context informatin about many items). > With this idea of tags, the incorrect data in the database would > always be incorrect because nobody is using it. The moderation system > currently allows somebody to correct that data, if they wish, whether it > is used or not. Then in the future if somebody wants to use it, it will > be correct. This is true for objective itemized data. Contextualizations are never 'correct' and things that do not get used do not get intertwingled into any contexts. > Remember, there are other people using the data (data feed customers) > where their users don't have direct interaction with the musicbrainz > site. And it's them I am thinking about. Tags are worthless if we leave them on the website. We can unleash the full power of tags only if we let them loose. And I mean: let them loose on the collections of people. Give people the ability to (1) get exiting tags for their artists/albums/tracks from MB, (2) to tag their music, (3) to *use* these tags with their players, and (4) to upload their tags to MB. Ideally 2 to 4 should be integrated into one step. I believe I have moved away from Shepard's original idea. Actually his mail has triggered me to tell my thoughts about tags in MB. These ideas have evolved over some time. Robert once told me that there is a fully coded folksonomy back-end lying around in some code-branch. So I have been thinking about the "front-end" (which means the interaction between the server and the users if you do not want to be MisunderstandingDonRedman). This is a much larger picture than to cluster albums from a box set. It will do that, too, though. DonRedman -- Words that are written in CamelCase refer to WikiPages: Visit http://wiki.musicbrainz.org/ the best MusicBrainz documentation around! :-) _______________________________________________ Musicbrainz-experts mailing list Musicbrainz-experts@... http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-experts |
|
|
Re: Clouds
by Rod Begbie
::
Rate this Message:
On 2/28/06, Steve Wyles <steve@...> wrote:
> But, musicbrainz is about recording the factual information for > the releases, subjectiveness doesn't come into it. *YES*! IMO, MB should never have "tags". Let last.fm deal with that -- their whole site is a community based around people and music. Let them work out all the interesting data around how people interact and feel about music. Let them have a tag like "seen live" which is useless on aggregate, but which has real meaning for an individual user. MB should stick to being the factual repository at the base of the tree. There are opportunities for cloudy-style data (My pet examples are a way to aggregate all CDs given free with Q Magazine, or all the releases in the "Now That's What I Call Music" series) but it has to be objective and voted-for. Rod. -- :: Rod Begbie :: http://groovymother.com/ :: _______________________________________________ Musicbrainz-experts mailing list Musicbrainz-experts@... http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-experts |
|
|
Re: Clouds
by Steve Wyles
::
Rate this Message:
On Tue, 28 Feb 2006, Don Redman wrote:
> > Robert once told me that there is a fully coded folksonomy back-end lying > around in some code-branch. So I have been thinking about the "front-end" > (which means the interaction between the server and the users if you do not > want to be MisunderstandingDonRedman). Okay, for this type of data, the tags idea is fine. But for factual information, it would be taking a backwards step. the example I used recently was: "Grass is green. I don't want to be told it is red because the other people looking at it recently were colour blind" Steve (inhouseuk) _______________________________________________ Musicbrainz-experts mailing list Musicbrainz-experts@... http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-experts |
|
|
Re: Clouds
by Nikki-12
::
Rate this Message:
On Tue, Feb 28, 2006 at 06:51:26PM +0100, Don Redman wrote:
> Yes, "'... rocken tun die andern!' is an album by 'gutfürnkeller'" is > either correct or wrong. This information is objective. It state a fact > about two items. > But whether gutfürnkeller's music is *called* 'Funk', or 'Jazz', or > 'nuJazz', or 'Blues' is not objective (people's oppinions or personal > classifications differ), nor does it state facts about an item. Rather it > provides *contextual information* for some items. Simon's example wasn't about genres though. I agree that genres are too subjective, and I also think that last.fm has a perfectly functional tagging system and I don't think that we need to recreate it. His example about discographies for a person should ideally be handled by relationships (and a better display of the information they hold). The other example, which is pretty similar to my original example, is not subjective data. The album either was or was not part of a 'collection'. Back to my example, Ghost in the Shell: Stand Alone Complex O.S.T. [1] was a soundtrack for Ghost in the Shell: Stand Alone Complex. Likewise, Eurovision Song Contest: Riga 2003 [2] was an album from the Eurovision Song Contest. 空耳ケーキ (Sora mimi keeki) / Raspberry heaven [3] was the single containing the opening and ending themes to the anime あずまんが大王 (Azumanga Daioh). None of those are subjective. My original thoughts envisioned something similar to the current artist pages, but they wouldn't be artists, I'll call them collections. 'Artists' like Disney, Love Hina, Nintendo, etc. would be collections. They'd contain albums (and artists and tracks, I guess, and possibly even other collections) which are related to the collection. Possibilities for collections would be things like films, TV shows, series, companies, etc. It's effectively saying "Here are two albums, they are linked together by a common thing: They're both ...", where the blank could be "soundtracks for the film ..." or "part of the series 'Now, That's What I Call Music'" or "freebies from ... magazine", etc. --Nikki [1] http://musicbrainz.org/album/ed86507f-eacd-4332-9e0d-0560376615df.html [2] http://musicbrainz.org/album/47a12760-6897-4809-b26c-032fa6f814be.html [3] http://musicbrainz.org/album/36deafd9-4960-487b-9b8a-ecd8925f6106.html _______________________________________________ Musicbrainz-experts mailing list Musicbrainz-experts@... http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-experts |
|
|
Re: Clouds
by DonRedman
::
Rate this Message:
On Wed, 01 Mar 2006 00:20:57 +0100, Nikki wrote:
> Simon's example wasn't about genres though. [...] > His example about discographies for a person should ideally be handled by > relationships (and a better display of the information they hold). > > The other example, which is pretty similar to my original example, is not > subjective data. The album either was or was not part of a 'collection'. > Back to my example, Ghost in the Shell: Stand Alone Complex O.S.T. [1] > was > a soundtrack for Ghost in the Shell: Stand Alone Complex. Likewise, > Eurovision Song Contest: Riga 2003 [2] was an album from the Eurovision > Song Contest. 空耳ケーキ (Sora mimi keeki) / Raspberry heaven [3] was > the > single containing the opening and ending themes to the anime あずまんが大 > 王 (Azumanga Daioh). None of those are subjective. > > My original thoughts envisioned something similar to the current artist > pages, but they wouldn't be artists, I'll call them collections. > 'Artists' > like Disney, Love Hina, Nintendo, etc. would be collections. They'd > contain > albums (and artists and tracks, I guess, and possibly even other > collections) which are related to the collection. Possibilities for > collections would be things like films, TV shows, series, companies, etc. > It's effectively saying "Here are two albums, they are linked together > by a > common thing: They're both ...", where the blank could be "soundtracks > for > the film ..." or "part of the series 'Now, That's What I Call Music'" or > "freebies from ... magazine", etc. Oh, but then it is simply an advanced entity with ARs, right? There could be a kind of grouping entity with a name (free form text) and a type box, movie, symphony, work, ...). And ARs would relate the elements to the type. Or would this need anything else? DonRedman -- Words that are written in CamelCase refer to WikiPages: Visit http://wiki.musicbrainz.org/ the best MusicBrainz documentation around! :-) _______________________________________________ Musicbrainz-experts mailing list Musicbrainz-experts@... http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-experts |
|
|
Re: Clouds
by DonRedman
::
Rate this Message:
On Tue, 28 Feb 2006 19:03:42 +0100, Steve Wyles wrote:
> "Grass is green. I don't want to be told it is red because the other > people looking at it recently were colour blind" But other people might contribute that grass is "soft", "fresh", "nice to have sex on" etc. and all that information is useful (unless you're single ;-P ). DonRedman -- Words that are written in CamelCase refer to WikiPages: Visit http://wiki.musicbrainz.org/ the best MusicBrainz documentation around! :-) _______________________________________________ Musicbrainz-experts mailing list Musicbrainz-experts@... http://lists.musicbrainz.org/mailman/listinfo/musicbrainz-experts |
| Free embeddable forum powered by Nabble | Forum Help |
