Closed philbarker closed 11 months ago
Thinking on the same topic from the other end: how might the Beatles song "Let It Be" be modelled as a record rather than sheet music and arrangements?
Other people will have a lot more expertise in applying FRBR to music, I know there is a whole discipline in music cataloguing about which I know nothing, so please forgive the naivety here.
This is only partial, and may be inaccurate but reflects the sort of thing that people who collect these things are interested in (at least judging by the cost and content of releases like the Anthology box set and Let It Be "Super Deluxe" release).
I think the significant difference between a Recording and a Release, which I think are two different types of Expression is interesting.
The composite nature of many of entities (each performer has their own part, multi-track recordings, albums etc.) might inform the relationships you need between entities.
There is also a fun possibility of traversing the WEMI hierarchy upward, in that a remix of a recording can (I think) have enough artistic expression in it to make it a distinct Expression (the original album and ...Naked may be of different Expressions).
Work: Let It Be
composer: Lennon & McCartney rights holder: Northern Songs, ATV, SONY...
Performed (Expressed) many times, some of which were recorded.
Recording 1: Run through
performers: Paul McCartney on guitar date: 19 September 1968
Recording n: a rehearsal
performers: Beatles (various instruments) location: Twickenham Studio date: 3 January 1969
Recording 27-A: a master recording
performers: Beatles + Billy Preston on Hammond Organ location: Apple Studio date: 31 January 1969 sound engineer: Glyn Johns
Recording 27-B: a master recording
performers: Beatles + Billy Preston on Hammond Organ location: Apple Studio date: 31 January 1969 sound engineer: Glyn Johns
These are multi-track recordings which can be mixed (i.e. different tracks included & mixes between different recordings, e.g. George Harrison over-dubbing his solo)
Recording-Mix1: "Single version"
based on: Recording 27-A producer George Harrison
Recording-Mix2: "Album Version"
based on: Recording 27-A producer: Phil Spector
Recording-Mix3: "Glyn Johns Version"
based on: Recording 27-A producer: Glyn Johns
Recording-Mix4 "Naked Version"
based on: Recording 27-A & Recording 27-B
And then releases, often as part of a release of a larger "composite" work, so lot's of interesting "has part" relationships between a Composite Expression an expressions of different works.
Release 1 as a single
release of: Recording-Mix1 date: 6 March 1970
Release 2 as part of "Let It Be, Movie"
release of: Recording 27-B date: ??? 1970
Release 3 as part of "Let It Be (Album)"
release of: Recording-Mix2: date: 8 May 1970
Release 4: as part of "Let It Be...naked (Album)"
release of: Recording-Mix4 date: 17 November 2003
Release 5: as part of "The Beatles: Get Back (TV series)"
release of: Recording 1
And then Releases can be re-pressed / re-issued sometimes in different formats, see 673 Versions of Release 3 on discogs.
as sold in record shops...
Wow @philbarker ! What this brings up to me is where the manifestation fits in vis-a-vis the expression. I presume that the manifestation would be the CD, DVD, or the TV show. Ross and I talked about how you model events, things that aren't fixed in a physical format that you can hold in your hand. A performance or a viewed TV show would be that, or a movie released in theaters.
I also think that a remix or even a different performance could be a new expression. If the same version is republished, say in a "best of" album, is that album a manifestation? We worked this week on whole/part and I will write that up for the group, but pop music albums are a good example of wholes with parts.
Then there is the question of whether someone would consider a version rendered by someone other than the Beatles to be a new work. We have defined high level "entity/entity" relationships, so this could be a work/work relationship.
What this brings up to me is where the manifestation fits in vis-a-vis the expression.
I found this on FRBR WEMI & Music: https://web.library.yale.edu/cataloging/music/frbr-wemi-music though I've no idea how widely accepted it is. It gives the musical score and performances as examples of Expressions and it says "Music can be performed, but only when it is recorded is there a manifestation" (which I think relates only to expressions as sound rather than printed scores)
Then there is the question of whether someone would consider a version rendered by someone other than the Beatles to be a new work.
That is difficult to judge. I was thinking about Dazed and Confused over the weekend. A change in genre is one example of a change in Work, and sure, the 3m46s-long folk song by Jake Holmes seems like a different work to the 30 minute long heavy rock version Led Zeppelin performed live: but when did it change? There is an incremental path through the Yardbirds cover version to the Led Zep extravaganza.
I guess there are different ways of copying or basing one musical work/expression on another
If the same version is republished, say in a "best of" album, is that album a manifestation?
I would say that part of it is. The album itself could be a "composite" Work (q: are poetry anthologies treated as [composite] Works in their own right?)
Just in case this isn't already too complicated, the Code4Lib journal this issue has an article about CD-R where someone can have taken a single session CD and added tracks from another session. Aside from the technical details (which are very technical) I am struck by the concept of "session" and think that may be the same as your "release" here, and confirms your division between a release and a recording.
flowchart TD
A[Let it Be] --> |expression|B(MasterA)
A --> |expression|P(Practice sesson)
A --> |expression|M(MasterB)
M --> |Manifestation|R("Let it Be" soundtrack)
B -->|subexpression - Johns| C{Mix1}
C -->|Manifestation| D[Album]
C -->|Manifestation| E[Single]
B --> |subexpression - Harrison|F{Mix2}
C --> |subexpression|G(Mix3)
F --> |subexpression|G
G --> |Manifestation|H(naked version)
This could be a vocabulary with:
flowchart TD
A[Let it Be] --> |expression|B(MasterA)
A --> |expression|P(Practice sesson)
A --> |expression|M(MasterB)
B --> |expression|C(Mix1)
B --> |expression|F(Mix2)
M --> |Manifestation|R("Let it Be" soundtrack)
C -->|Manifestation| D[Album]
C -->|Manifestation| E[Single]
F --> |relatedExpression|G(Mix3)
C --> |relatedExpression|G(Mix3)
G --> |Manifestation|H(naked version)
This could be a vocabulary with:
flowchart TD
A[Let it Be] --> |expression|B(MasterA)
A --> |expression|P(Practice sesson)
A --> |expression|M(MasterB)
B --> |related expression|C(Mix1)
B --> |related expression|F(Mix2)
F --> |relatedExpression|G(Mix3)
C --> |relatedExpression|G(Mix3)
M --> |Manifestation|R("Let it Be" soundtrack)
C -->|Manifestation| D[Album]
C -->|Manifestation| E[Single]
G --> |Manifestation|H(naked version)
This could be a vocabulary with:
flowchart TD
A[Let it Be] --> |session|B(MasterA)
A --> |sesson|P(Practice sesson)
A --> |session|M(MasterB)
B --> |mix|C(Mix1)
B --> |mix|F(Mix2)
F --> |mix|G(Mix3)
C --> |mix|G(Mix3)
M --> |product|R("Let it Be" soundtrack)
C -->|product| D[Album]
C -->|product| E[Single]
G --> |product|H(naked version)
This uses properties that would be associated with openWEMI: session = expression mix = expression based on session (there may be the need for a "re-mix" here) product = is physical output
Just filling in with some knowledge of music industry and recordings here. The term music recording master can be problematic in it's current use since the term is somewhat established (yet vague) in the recording process. There already is a concept of "master tapes", which depending on context usually means the original recorded multitracks which are later mixed. Sometimes it refers to the mixed 2-track which is copied for distribution. Which jacks into the term master which is the product of the process of mastering optimizing the mixed tracks for a particular media or distribution. Which means there can be different masters for the CD, Vinyl and digital streaming service, often but not always done by the same mastering engineer. A new medium can trigger remasters years later.
So I think we need to be very precise what we mean by master here. I have seen services like discogs use the term "master release" but in the context it is quite evident that it is the common description between each released format but I am not so sure the difference between what is common and what is the original release is particulary widely upheld (not too familiar with their descriptive rules). Example of master release https://www.discogs.com/master/24212-The-Beatles-Let-It-Be
We can definitely change the terminology here, since this is just to be an example. I can imagine that there can be different "recordings" along the workflow, with one being pretty much the original that happened in the studio. To me, the main thing here is that the ability to mix and re-mix leads to a variety of expressions, some of which have relationships to each other. Whether or not one wants to specify the exact relationships depends on the use case.
If anyone has examples that show different points of view from what we have here that would be very useful. Thanks.
I think the idea of sessions makes sense as input to whatever comes next in the production process. Altough each session can contain several takes of the same song (each take would be considered an expression). Also for a remix there can be additional sessions. A good example of this is https://en.wikipedia.org/wiki/A_Little_Less_Conversation where the original session had two takes which was released independently as album (take 10) and single + soundtrack (take 16). Another suspected re-recording later known as take 2 and then the remix by Junkie XL.
I have given some further thought into this and read some of the previous discussions while trying to reconcile it with a use-case of adding an ISRC-identifier to a recording. The assignment of this identifier is not always a clear cut case but for the purpose of this example I am doing this in the role of a record label. For reference and further reading I refer to ISRC Handbook - (https://creativecommons.org/licenses/by-nd/4.0/)
International Standard Recording Code (ISRC) Handbook – 4th Edition, 2021, International ISRC Registration Authority – Page 10 states:
"The key principles for ISRC assignment are (a) that each distinct recording shall be assigned one and only one ISRC and (b) that any particular ISRC shall be assigned to one and only one recording."
and the model described is very closed to what has been already discussed:
The recording
is also defined as an expression derived from the indecs framework while not being recent in any way had an ambition to make metadata-systems for media products interoperable with models such as FRBR/CRM/DC with the Expression
being modeled as an event of creation:
8.2.2 "An expression is a performance – an event which is in itself regarded as a creation and may or may not be recorded or fixed in a manifestation, and may be reproduced by some form of recording playback. The expression is the event which is recorded, not the physical or digital recording itself, which is a manifestation. One expression may be recorded and copied onto many media while maintaining its integrity. An expression may give rise to an abstract work; at the same time it may be an expression of an existing abstract work."
So having come back full circle, the things to be identified by an ISRC can only be a sound recording
or a music video recording
never the product
. The ISRC should then be encoded in the formats produced such as the Mp3-file, CD, DVD regardless of distribution. For example An album CD of 10 tracks assigns 10 ISRC codes, one to each of them. "the carrier shall (when technically feasible) include the ISRC of each recording that it contains."
Like many other assignable identifiers a recording with exploited parts (such as a symphony) may be assigned an ISRC as a whole and each of its components may also be assigned an ISRC, but I think we do well to stay out of more complex cases for now.
For an ISRC-code to be applied it should follow some guidelines on attributing and maintaining metadata:
Which transformative events renders new identifiable recordings and which do not are defined and can be found in the above referenced handbook:
A.8 NEW ISRC NEVER REQUIRED
A new ISRC shall never be assigned in the following circumstances:
A.8.1 A recording is encoded using a different technology
A.8.2 A high-resolution version of a recording is created
A.8.3 A version for a different business model is created
A.8.4 A compilation is created Where the recording is reused on a compilation without change, re-mixing or re-mastering, the same ISRC shall be used.
A.8.5 A recording is licensed from a third party with its ISRC
A.9 NEW ISRC ALWAYS REQUIRED
A new ISRC shall always be assigned in the following circumstances:
A.9.1 A live version of a studio recording is released
A.9.2 A pre-release version is created for promotional purposes If the pre-release version is different from the final release and it is to be used outside the Registrant’s control, a new ISRC shall be assigned.
A.9.3 Different versions are created using different studio ‘takes’
A.9.4 An edited version is created
A.9.5 A ‘cover version’ is created
A.9.6 A music video is created A music video is always different from an audio recording and it shall not be identified by the ISRC of the audio recording on which it is based.
A.9.7 Different music video versions are created
A.9.8 A remixed version of a recording is created
A.9.9 A different version is created by adding further tracks to a recording
A.9.10 An extended version is created
A.9.11 A clip is taken from the recording for promotional use or as a ringtone
A.9.12 An interlude, skit or interview is created
A.9.13 A callout is created If the callout (a clip from a recording which includes the “hook”) may be exploited separately (rather than just being used within the recording), an ISRC shall be assigned.
Taking the earlier mentioned example of "A little less conversation" and considering how many versions there are, this pruned and curated ttl-example is a bit fictious and solely based on information from Wikipedia/Discogs/Musicbrainz (Sorry all Elvis fans!)
# Manifestations
<JXL-Single1> a :Product ;
:title "Elvis Vs JXL – A Little Less Conversation" ;
:releasedBy "RCA" ;
:featuring [ <JXL-Remix1>, <JXL-Remix2>, <Elvis-Recording1-TakeX-EditX> ] ;
:masteredBy "Drew Lavyne, (tracks: 1, 2)" .
<Elvis-Compilation1> a :Product ;
:title "ELV1S 30 #1 Hits" ;
:artist "Elvis Presley" ;
:releasedBy "RCA Victor" ;
:featuring [ <JXL-Remix1> ] .
<Elvis-Compilation2> a :Product ;
:title "Essential Elvis Presley" ;
:artist "Elvis Presley" ;
:releasedBy "RCA Victor" ;
:featuring [ <Elvis-Recording1-Take16> ] .
# Expressions
<JXL-Remix1> a :SoundRecording ;
:artist "Elvis vs JXL" ;
:title "A Little Less Conversation" ;
:alternative "(Radio Edit Remix)" ;
:duration "3:30" ;
:date "2002" ;
:ISRC "USRC10200288" ;
:expresses <Song1> .
<JXL-Remix2> a :SoundRecording ;
:artist "Elvis vs JXL" ;
:title "A Little Less Conversation" ;
:alternative "(12" Extended Remix)" ;
:duration "6:07" ;
:date "2002" ;
:ISRC "USRC10200293" ;
:expresses <Song1> .
<Elvis-Recording1-TakeX-EditX> a :SoundRecording ;
:artist "Elvis Presley" ;
:title "A Little Less Conversation" ;
:duration "1:39" ;
:date "2002"
:ISRC "USRC10200294" ;
:expresses <Song1> .
<Elvis-Recording1-Take16> a :SoundRecording ;
:artist "Elvis Presley" ;
:title "A Little Less Conversation" ;
:duration "2:00" ;
:date "1968" ;
:ISRC "USRC19900989" ;
:expresses <Song1> .
# Work
<Song1> a :MusicalWork ;
:title "A Little Less Conversation" ;
:writtenBy ["Billy Strange", "Mac Davis"] ;
:ISWC "T-071.172.457-0" .
There are interesting aspects in the documentation how to handle different lengths of the recording based and whether edits etc. is done on purpose.
Not much information could be found on mixing and mastering personnel regarding these recordings however remastering
has an interesting section regarding the artistic purpose:
"Re-mastering is a broad term that covers many different processes. A new ISRC shall be assigned if (and only if) the processes applied to a recording during re-mastering involve the application of creative input to the recording itself."
with several examples of which operations would not be considered a new recording.
All in all if there is uncertainty the guiding principle comes down to purpose and artistic intent whether something is considered a new recording. In practicality and my experience it is the same with any manual Identifier assignment, new ones get assigned where not needed and others get reused where they shouldn't and some just get it plain wrong. However the framework around it seem to map well to open-wemi, but would perhaps require more diverse examples and a comparison to other descriptive rule-sets.
Thank you @klngwll for this. I would like to make it into an example in the examples folder. I'll try to add a diagram. My goal is to show various interpretations that use WEMI in different ways so that it is clear that we aren't trying to enforce a single view but encourage developers to bend WEMI to their own needs, within reason.
flowchart LR
A[Song1] --> |SoundRecording|B(Elvis-Recording1-Take16)
A --> |SoundRecording|C(Elvis-Recording1-TakeX-EditX)
A --> |SoundRecording|D(JXL-Remix2)
A --> |SoundRecording|E(JXL-Remix1)
B --> |product|F(Elvis-Compilation2)
E --> |product|G(Elvis-Compilation1)
C --> |product|P(JXL-Single1)
D --> |product|P
E --> |product|P
Some articles that might provide info for this (albeit from a library cataloging point of view):
Smiraglia, Richard P. "Musical Works as Information Retrieval Entities: Epistemological Perspectives." In ISMIR, pp. 85-91. 2001. https://www.academia.edu/download/36385388/smiraglia.pdf
Smiraglia, Richard P. "Musical works and information retrieval." Notes 58, no. 4 (2002): 747-764.
Riley, Jenn. "Application of the Functional Requirements for Bibliographic Records (FRBR) to Music." In ISMIR, pp. 439-444. 2008. https://jennriley.com/presentations/ismir2008/riley.pdf
@philbarker DRY. :-) Just align with MO - Music Ontology - http://purl.org/ontology/mo/ , where Yves Raimond and team have already done most of the work for most of the world. Look at their Getting Started page...there's an entire Production Workflow example that will wet your appetite. So there just needs to be a mapping between WEMI and MO. That mapping can eventually just live in Wikidata.org, like I did with Schema.org. So happy to help add SKOS relations and mapping into Wikidata.org for MO, once your namespace is official? (or the mapping can live in other places/files also)
@thadguidry thanks :-) This issue lead to a couple of things in the primer, one showing how to reference openWEMI when creating your own ontology, the other how to use it in instance data when using an existing ontology that didn't reference openWEMI. Obviously we can't use an existing ontology for the first case. We used schema.org for the example with an existing ontology as it's widely used. There is a stub in the wiki that will build on those examples, that's probably where MO would fit best.
@kcoyle do you think this issue has served its purpose and can be closed?
@philbarker thanks, I do think this no longer needs to be an issue, but I'd like to encourage some more examples since the music environment seems to be a rich source. @thadguidry if you have some examples of MO instance data we could start to work with that. I believe you can add them to the Wiki and we can work with them there. The example that we have is for recorded music, so some other type of music resource would be good.
Is anyone familiar with Music Brainz? They have some WEMI-like entity / classes (Work, Recording, Release [they don't really deal with Items, I guess]) and a table of relationships between them.
Might be possible to do a music example from this?