» Dezrann corpus/developer documentation

Specifying corpus and piece data and metadata

A metadata/corpus.json file gathers all data/metadata information on pieces into a corpus, that is:

corpus (corpus metadata)
for each piece in the corpus, opus (piece metadata) and sources

You are welcome to discuss or to propose MR to improve the definition and the handling of such metadata.

{
  "corpus": {
    // Corpus metadata [mandatory: at least id and title]
    "id": "bach-fugues",
    "title": "Das wohltemperierte Klavier, Buch I",
    (...)
  },

  "pieces": {
    "bwv846": {
      // Piece metadata [mandatory, see below]
      "opus": { ... },

      // Sources (scores, audio, analyses, ...)
      // [mandatory: at least one score/audio source]
      "sources": [ ... ],

      // Piece settings
      "settings": { ... }
    },
    
    "bwv847": { 
      ...
    }
  },

  "settings": {
      // Optional
      "access": "public"
  },

  "template": {
      // Optional
      ...
   }
}

Files should be referred by a URL (preferably to some git repository) or a stable external identifier. They can also be referred by a local path (relative to the corpus.json file).

There can be optional settings specific to Dezrann or any other application.

The metadata/corpus.json file can be split in several files, for example when the corpus section is hand-curated whereas the pieces are produced by another script.

Examples

The /metadata directory contains several examples, such as:

minimal.json (minimal example, two pieces, scores, audios, analyses)
collabscore-saintsaens.json
openscore-lieder-corpus.json (corpus) and openscore-lieder.json (pieces, generated from data in another git)
bach-fugues.json (with some templating yielding bach-fugues.full.json)
mozart-piano-sonatas.json (with some templating yielding mozart-piano-sonatas.full.json)
supra.json and weimar-jazz.json (only corpus, the upload of the pieces is directly handled by custom scripts)

For most of the baroque/classical/romantic music, the score is the reference, there may be recordings. For jazz/pop, a recording is the reference, there may be a score, such as in several examples on edmus.json.

`corpus` Metadata

This section describes the corpus as a whole. This is used both on http://dezrann.net/corpora and on corpus pages such as http://dezrann.net/explore/mozart-piano-sonatas.

To draft a corpus, only id and title are mandatory. The other fields will have to be carefully written/reviewed/translated to prepare the release of a public corpus.

id: Mandatory, unique identifier, such as "bach-fugues", "corelli-trio-sonatas" or "schubert-winterreise.

🏳️ shorttitle: Title with < 25 characters, to be displayed on http://dezrann.net/corpora and at other places with reduced space

🏳️ motto: Short text (< 120 characters, nominal sentence) advertising the corpus, including some number of the works

image: url of one image (recommended size: XXXpx XXXpx) to illustrate the corpus

🏳️ title: Mandatory, title

🏳️ text: 3-6 lines presenting the corpus, with a historical / musicological perspective (and not referring to Dezrann). Links (in markdown) to Wikipedia or other sites are welcome.

🏳️ availability: (for the general public) 1-4 lines detailing what is actually in Dezrann, including stating which content is there (scores, autographs, analyses, synchronized audio), both qualitatively and quantitatively.

status: (for the technical audience) short message summarizing the main issues with this corpus, that will be displayed on https://doc.dezrann.net/status

🏳️ contributors.*: see metadata.md#the-contributors-bloc. Before a public release, the corpus should contain a contributors.maintainer field.

showcase: a list of 1-4 ids of pieces in the corpus with a particularly high quality (availability of sources, quality of those). They will be randomly displayed/showcased from some places such as http://dezrann.net/corpora

Optional external references ref*: see below

quality:corpus and quality:corpus:metadata: See quality.md. If you just completed a draft of corpus.json, start with these both fields to 1.

Other fields (optional)

genre

opus: catalogue numbers or opus numbers

`piece` Metadata and Data

The piece data may be either static / fully redacted, or produced by a script from some, and/or maintained with templates, see templating.md.

`sources` Data

At least one source has to be provided.

 "sources": [
  // Scores (.MEI (preferred), .musicxml, .mscz, .krn), see scores.md
  {
    "score": "http://gitlab.com/bla/bli/02.krn",
    "license": "CC-BY-SA-4.0",
    "contributors": {
      "encoder": "Jane Foo",
      "editor": "KernScores"
    }
  },
  {
    "score": "http://bla.net/sonata-12.mscz",
    "measure-map": "http://bla.net/sonata-12.mm.json"
  },
  {
    "score": "vivaldi-04.mei"
  },
  // Scores from Neuma pipeline
  {
    "score:neuma": "all:collabscore:saintsaens-ref:C080_0"
  },
  {
    "score:gallica": "ark:/12148/bpt6k1162028r"
  },

  // Audio/video sources
  {
    "audio":  "http://gitlab.com/bla/bli/bla-07.mp3",
    "source":  "http://a-wonderful-open-music-project.org/",
    "license": "CC0-1.0"
  },
  {
    "video":  "http://a-wonderful-open-music-project.org/my-video-07.mp4",,
    "source":  "http://a-wonderful-open-music-project.org/",
    "license": "CC0-1.0",
     "contributors": {
      "performer": "Clara Dee"
    },
    "info": "Studio recording"
  }
  {
    "video:yt":  "df6DFfs",
    "contributors": {
      "performer": "Clara Dee"
    },
    "info": "Live recording at the Schnupz Concert Hall"
  },
  {
    "audio:yt":  "df6DFfs",
  },
  {
    "audio:yt":  "df6edfs",
    "synchro": "http://gitlab.com/bla/bli/synchro.json"
  },

  // Analyses that will be displayed in Dezrann
  {
    "analysis": "https://gitlab.com/foo/bar/haydn-symph099-mvt1.dez",
    "contributors": {
      "analyst": "Dinh-Viet-Toan Le, Francesco Maccarini"
    },
    "ref:doi": "10.1145/3543882.3543884"
  },

  // Special sources, with provided image(s) and position files
  // Any scan, with positions
  {
    "images": [ "http://bli/scan-07-page1.jpg", "http://bli/scan-07-page2.jpg" ],
    "positions": "positions-07.json"
  },
  {
    "images": [ "http://bli/scan-07.jpg" ],
    "positions": "positions-07.json"
  },

  // Scan of piano rolls (to be better specified)
  {
    "audio": "bla/bli07.mp3",
    "image": "bli/roll-07.jpg",
    "positions": "positions-07.json"
  },
  {
    "grid": "| D A | Bm D/F# | G D | Em7 A7"
  }
  ]

Source: Basic information

For each source, one (and only one) of these fields has to be provided:

score (url or file)
- or possibly through an identifier score:neuma or score:gallica
- possibly with a measure map
video (url or file)
- or possibly through an identifier video:yt
audio (url or file), including videos that are… not real videos
- or possibly through an identifier audio:yt
analysis (url or file in .dez format)

Source: Metadata

These fields are optional

🏳️ contributors (dictionary, see below)
🏳️ name (short name used to refer to this source and distinguish it from others. For example, for a score, it may correspond to contributors.encoder. For an audio, it is usually the same as contributors.performer or contributors.artist. But it can also be contributors.editor when relevant, for example to acknowledge a open-data project)
🏳️ info (short string, < 100 characters)
license (SPDX identifier)
source (url)
album (string)

(Jazz: Concert place / Recording Label, to be detailed)

`opus` Piece Metadata

`opus`: Basic information

(when applicable) corpus or collection

🏳️ Use corpus for generic names such as "Mozart piano sonatas" (they do not bring more information than what is in piece:title and composer)
🏳️ Use collection for data giving more information ("London symphonies", "Le quattro stagioni"). The two fields may be different.

(mandatory) id: Unique identifier, such as "bach/bwv847". Should include some opus information and be consistent with other identifiers used in Dezrann. (Note that there is also a id field outside opus, it will be removed at some time.)

`opus`: Titles

(mandatory) 🏳️ title or piece:title: We follow the Open Opus style guide, except that we do not put the opus information (see opus below).

(when applicable) 🏳️ nickname or piece:nickname

(when applicable) movement:num and movement:title (in this case, we do not use title and nickname but rather piece:title and piece:nickname)

(mandatory) opus: catalogue number or opus number, such as "opus": "K.551". There are something numbers that are outside catalogue/opus numbers and that may be encoded as other fields, such as "symphony:num": "41" for Mozart’s Jupiter.

`opus`: Other information

genre

key: key/tonality of this file (movement, not piece), such as D minor of Bb major

meter: such as 4/4 or 3/4, 6/8. More complex meters can be given through measure maps in the score source.

Common fields for `corpus`, `opus`, and one of the `sources`

Basic information

year: year, or range of years

`ref`: External references

{
  "ref": "https://github.com/DCMLab/mozart_piano_sonatas/tree/main/scores",
  "ref:gallica": "ark:/12148/btv1b55002567w",
  "ref:neuma": "all:collabscore:saintsaens-ref:C080_0",
  "ref:doi": "10.1145/3543882.3543884",
  "ref:wikipedia": "Symphony_No._35_(Mozart)",
  "ref:musicbrainz": "611a55ef-cfb4-3bbf-9d65-7d4af5506093",
  "ref:musicbrainz:recording": "ed80fa40-4871-4669-9509-18bcfee420fd",
  "ref:imslp": "19_Sonatas_for_the_Piano_(Mozart,_Wolfgang_Amadeus)",
  "ref:kernscores": "mozart/sonata/sonata04-1.krn",
  "ref:rism": "990062490",
  "ref:wjd": 70
}

ref is any URL. The other ref:s are identifiers on some sites/databases, as defined on RefsBase.ts. When possible, put a reference of the particular movement. But it’s also acceptable to put an external reference to a piece or even to the collection.

It is strongly recommended to include as much information as possible. Specifically, for scores, it is advisable to include ref:rism to trace the primary source used by the encoder(s).

The `contributors` bloc

🏳️ One of the three following fields is usually defined:

(for opus)

composer
artist (on pop music, when the music is frequently related to the artist rather than the composer)
performer (for transcription of jazz solos)

(for an audio source)

performer

🏳️ The following fields can also be used:

lyricist
arranger
editor (edition, publisher)
encoder (digital encoding of an existing edition)
transcriber (transcription of solos)
analyst (supervision of analyses)
annotator (annotation, following precise analytical guidelines)
maintainer (long-term maintainance of the corpus or the piece)

Localization

In either corpus, piece, or source metadata, the 🏳️ fields can be localized, such as:

    "collection": "Le quattro stagioni",
    "collection:en": "The four seasons",
    "collection:fr": "Les quatre saisons",

    (...)

    "ref:wikipedia:fr": "Les_Quatre_Saisons",

Do not put :xx for the title in the original language.

Note that even contributor names such as composers may be known with some variations across languages.

"contributors": {
    "composer": "Johann Sebastian Bach",
    "composer:fr": "Jean-Sébastien Bach"
}

Do not localize numeric or formalized fields such as key, opus, or genre.

Examples

corpus Metadata

Other fields (optional)

piece Metadata and Data

sources Data

Source: Basic information

Source: Metadata

opus Piece Metadata

opus: Basic information

opus: Titles

opus: Other information

Common fields for corpus, opus, and one of the sources