Descriptive Metadata in the Music Industry: Why It Is Broken And How to Fix It—Part Two
"Why hasn't this been fixed yet?"
- Business case not sufficiently made yet. Findings that content-enriched records increase selection recent, maybe not well known, or because studies concern libraries, may have been disregarded as irrelevant to music industry.
- People may not want to increase filesize by adding metadata. Not a reasonable fear, metadata tiny relative to encoded media.
- Lack of track-level data, lack of abstracted data models.
- Challenge of offering metadata-enhanced discovery while maintaining clean UI.
- Determining validity of identity; too little oversight means inaccuracies, too much is impossible to deal with volume. (MusicBrainz offers reasonable compromise: edits must meet detailed requirements, be voted on by other editors; only 227 of 700k editors have earned "auto-edit" capability through proven trustworthiness; transparency and ability of any interested stakeholder to make corrections.)
- Stakeholders may have too much value in proprietary silos to be interested in collaboration (eg Gracenote sold in 2014 for $170m).
- Control over metadata old paradigm, mindset being slowly replaced
Globally Unique Abstracted Persistent Identifier
- Must be able to operate at multiple levels of abstraction (see part 1); some current systems offer limited/hardcoded level of abstraction through hierarchy in identifier eg country code in ISRC
- A truly forward thinking GUAPI could also address location of time within media
- ISRC most widely used, primarily track record, only years 1940-2039 allowed. "After decades of development, ISRC still lacks even a simple reference database of descriptive metadata that can be queried for distinction between similar entities and for conflict resolution"
- MusicBrainz ID (MBID) a 36 character UUID, randomly generated, not hierarchical, and abstracted/applied to works, releases, and also artists and labels, and mapped to other schemas, in the MusicBrainz database
- GRid a complement to ISRC, but just one level of abstraction, not widely adopted yet, not coupled with database
- GRD being developed as an extremely complicated rights management metadata database, not developing a descriptive metadata database
- Conceptually ISWC:ISRC is 1:many, but there is no field in ISRC registry for ISWC
- ISNI and IPI assigned identifiers for various kinds of contributors and publishers
- Proprietary identifiers are likely to not be truly persistent (example of label switching distributor, reassignment of identifiers in iTunes store)
- Lack of standards indicating how ISRC to be embedded in files
- Developers should start using MBID now, meets requirements (see part 1) and can be mapped to other identifiers
- Separate identifiers for various abstraction levels could be easier to manage; but the current systems have proven very difficult to manage, surely coordinated system would be better
- Function specificity could be essential to persistence -- just one level of abstraction needing fixing could break abstracted system; but the current non-abstracted system is already broken
Components of solution:
- Rally will to cooperate. Proposes study on richer metadata/abstract model spurring sales.
- Develop schema for descriptive metadata through standards organization such as DDEX or ISO. Strongest candidates are MusicBrainz and CCD schemas, but they are complementary, need to be combined to be comprehensive.
- Develop GUAPI. Path 1: ISRC, ISWC, and GRid merge and develop global abstracted identifier system. Path 2: New effort to develop system.
"If MusicBrainz is to be the source of a standardized schema, they may need a companion organization to advocate for them within the music industry." Suggests The Recording Academy.