Descriptors
Descriptors indicate characteristics of an artist, album edition, or recording to help drive use cases such as catalog organization, discovery, and analytics. There are currently eight descriptor types:
- Artist-level Descriptors: Genre, Language, Origin, Era, Artist Type
- Recording-level Descriptors: Sonic Mood, Sonic Style, and (Perceived) Tempo (hereafter referred to as Tempo).
Descriptor Value
A weighted descriptive attribute of an Artist or Recording. Gracenote editorial staff assign Artist-level Descriptors. Machine listening processes trained from the same expert editorial knowledge assign Recording-level Descriptors.
- There can be up to 10 different descriptors of the same type assigned to a single Artist, Album or Recording endpoint.
- For Genre, Language, Origin, Era, Artist Type, and Tempo, the weights of all descriptors of the same type assigned to a single Recording or Artist entity will add up to approximately 100.
- For Sonic Mood and Sonic Style, the sum of the descriptor weights may sometimes add up to less than 100, due to the full vector extending beyond the 10 most relevant positions published in the GMD API.
- For a given entity, the descriptor with highest weight for each descriptor type is considered the “primary descriptor” for that type.
Genre
Genre describes the musical style of the Artist, Album, or Recording. Each Artist, Album and Recording may have one or more Genres assigned, each with a weight associated with it, indicating its relative relevance for that entity. Gracenote’s international team of editors select from a controlled vocabulary of Genre descriptors values in making these assignments. The combination of genre descriptor values and their weights that are assigned to an Artist, Album, or Recording are referred to as their “Genre Profile”. A specific Album or Recording's primary Genre may be different from that of its album or recording Artist.
Language
Artist Language indicates the languages in which the artist has recorded in. Each artist may have one or more languages assigned, each with a weight associated with it that indicates its relevance for that artist.
Origin
Origin indicates the geographic region(s) with which the recording artist is most strongly associated. Note that this is not necessarily their birthplace or country, but rather the music scene with which they are most strongly linked. Most Artists have just a single origin assigned, though some do have more than one.
Era
Era indicates the time period(s) in which the artist created its most relevant recordings. This is editorially determined based on a combination of when the Artist recorded most of their work, and when, within that range, the Artist’s best-known and most critically – acclaimed Recordings were made. Era profiles include weight assignments, which indicate the relative importance of different periods within the Artist’s career. This enables accurate description of Artists whose active recording period spanned multiple periods.
Artist Type
Indicates the configuration in which the artist typically performs. This is described in terms of Artist Grouping (Solo, Duo, Group), Artist Gender Composition (Male, Female, Non-Binary, Mixed), Instrumentation (Vocal, Instrumental), Vocal Grouping (Solo, Duo, Group), and Vocal Gender Composition (Male, Female, Non-Binary, Mixed).
Sonic Mood
For each Recording, the (up to) ten strongest Moods detected (out of the 400+ Sonic Mood classes) are included in the GMD API. Depending on how many different significant Moods are detected in the Recording, 1 to 10 Mood Descriptor Values will be provided. Each one is assigned a weight indicating how much out of a total of 100 points is attributable to that Mood for the recording.
The “Primary Mood” is the most prominent Mood detected across the duration of the Recording. This determination is made according to the relative strengths of each Mood detected through supervised machine listening and heuristics.
Sonic Style
Sonic Style describes musical style elements detected in a Recording using machine learning, based purely on the nature of the recording audio itself, and are different from Artist Genres, which describe the musical and cultural context of an artist and are assigned editorially.
Tempo
Tempo indicates the 3 most perceptually relevant beat rates detected in the Recording. Each rate also has a weight associated with it which indicates its relative relevance across the curation of the Recording. Tempo Descriptor Values correspond to beat rates from 0 to 300+ BPM in 10s. The “Primary Tempo” is the most prominent Tempo, as determined through analysis and heuristics.
Album Release Type
Release Type applies to Albums and indicates whether it is an original release or a single artist’s greatest hits, various artists compilation, or other release type. This field can be used for filtering and ordering, for example, to prefer the track instances of a desired Recording on a Main Canon Album over alternate track instances of the same source Recording on a Various Artists Compilation.
| Release Type | Description |
|---|---|
| Main Canon Collection | An original album by multiple artists Example: Saturday Night Fever Soundtrack |
| Single / EP | A commercially available Single release Example: Borderline [Maxi Single] |
Descriptor Hierarchical Data Structures
Descriptor Hierarchies enable the presentation of the information contained in the Descriptor Values in simpler, more user-friendly display categories. The Descriptor Hierarchies are not used for calculating artist or recording similarity, but rather for categorical display, organization, and navigation in the end-user UX.
There is a single Descriptor Display Hierarchy for each of the following Descriptor Types: Origin, Era, Artist Type, Sonic Mood, Sonic Style, and Tempo. For Genre, however, there are a total of 17 Hierarchies. This is because to provide the best user experience across the globe and each application use case, Gracenote provides regionally appropriate genre hierarchies for each of 9 international regions - and for each of these regions except for India, which only has a Simplified Hierarchy, a Simplified and a Detailed version is available.
Additionally, all display category hierarchy nodes are localized and translated into over thirty languages.
It is important to keep in mind that regardless of which of the 17 Genre Hierarchies is presented to an end user, and which localization strings are displayed, all source data at the Descriptor Value level utilizes the same controlled vocabulary. That is, there is only a single set of Descriptor Values used by Gracenote editors and machine listening systems worldwide, and all annotations are made using this single controlled vocabulary. This ensures that the semantic integrity of all descriptive data is always maintained globally.