It is useful to have some common, basic patterns to follow when using a very open model and ontology. The following patterns have been agreed on as useful ways to think about our cultural data, through working with objects, other entities and collections from across many different museums and organizations. They are foundational to all of Linked Art, and used extensively both to describe the core entities of interest (objects, works, people, places and so on) as well as within the descriptions (for example to give a label to a description, or a classification to a span of time).
These patterns are presented below with examples of how they are used in practice, but these are not intended to be exhaustive. The documentation for the different entity types will include more information about how they are used in different circumstances.
Core Properties
There are a few core properties that every entity must have for it to be a useful part of the world of Linked Art and Linked Open Usable Data (LOUD):
@context
is a reference to the context mapping which determines how to interpret the JSON as LOUD. It is not a property of the entity being described, but of the document. It MUST be present.id
gives the URI that identifies the entity. Every core entity MUST have exactly one URI.type
gives the class of the entity (orrdf:type
in RDF). Every entity MUST have exactly one class. This allows software to align the data model with an internal, object oriented class based implementation. Classes determine which properties or relationships may be associated with the entity._label
gives a human readable label as a string, intended only for developers or other people reading the data to understand what they are looking at. Every entity SHOULD have exactly one label, and must not have more than one. It is just a string, and does not have a language associated with it -- if multiple languages are available for the content, then implementations can choose which is most likely to be valuable for a developer looking at the data.
The classes used for the core entities we describe are summarized as below, with links to the full documentation about them:
Class Documentation Description HumanMadeObject
Objects Physical things you can touch, e.g. a painting DigitalObject
Digital Digital things that exist only in computers, e.g. a web page Person
Agents A human person, e.g. Rembrandt Group
Agents A group of people, capable of collective action, e.g. a museum department Place
Places A place, typically somewhere on earth that could have coordinates, e.g. Paris VisualItem
Visual Work Conceptual image/visual content shown by a physical or digital object, e.g. the image we know as "The Mona Lisa", as opposed to the physical painting LinguisticObject
Documents Textual content carried by a physical or digital object, e.g. the text of The Lord of the Rings PropositionalObject
Exhibitions An entirely abstract work that is neither linguistic nor visual, e.g. the notion of an exhibition or piece of performance art Type
Concepts A category, concept, or similar, e.g. centimeters or sculpting Set
Collection A set or collection of some number of other entities, e.g. a museum collection Activity
Provenance, Exhibitions An activity carried out by people or groups, e.g. an exhibition or the transfer of ownership of an object between parties
Example:
The simplest possible object description of "Spring" has a URI, a class and a label.
{
"@context": "https://linked.art/ns/v1/linked-art.json",
"id": "https://linked.art/example/object/spring/1",
"type": "HumanMadeObject",
"_label": "Jeanne (Spring) by Manet"
}
Types and Classifications
CIDOC-CRM, our underlying ontology, is a framework that must be extended via additional vocabularies and ontologies to be useful. The provided mechanism for doing this is the classified_as
property, which refers to a term from a controlled vocabulary. This is in contrast to the type
property described above, which is used for a limited number of core, ontological classes. The classified_as
property is thus a way to be more specific about the sort of entity, while maintaining the core information as the class using type
. Controlled vocabulary entries should not be used with type
, nor classes used with classified_as
. More than one classification can be associated with each entity, even classifications of the same general sort (such as multiple type of object classifications, or multiple nationalities for a person).
While any external vocabulary of terms can be used, the Getty's Art and Architecture Thesaurus is used whenever possible for consistency and that it is already widespread in the museum domain. The set of terms that have been identified as useful are listed in the vocabulary lists for requirements when a particular choice is essential for interoperability and recommendations for shared semantics.
Core entities (those listed above) SHOULD have at least one classification, MAY have more than one or MAY have none.
Use cases for this pattern are in almost every example, but include:
- The object being an art object
- The type of an art object being a painting
- The type of a description being a materials statement
- The type of an identifier being an accession number
- The type of an organization being a museum
- The type of a place being a gallery
Example:
The classification of "Spring" (an instance of the class HumanMadeObject
) is a painting (aat:300033618), and an artwork (aat:300133025):
{
"@context": "https://linked.art/ns/v1/linked-art.json",
"id": "https://linked.art/example/object/spring/2",
"type": "HumanMadeObject",
"_label": "Jeanne (Spring) by Manet",
"classified_as": [
{
"id": "http://vocab.getty.edu/aat/300133025",
"type": "Type",
"_label": "Artwork"
},
{
"id": "http://vocab.getty.edu/aat/300033618",
"type": "Type",
"_label": "Painting"
}
]
}
The classification of "Paris" (an instance of the class Place
) is a city (aat:00008389):
{
"@context": "https://linked.art/ns/v1/linked-art.json",
"id": "https://linked.art/example/place/paris/1",
"type": "Place",
"_label": "Paris",
"classified_as": [
{
"id": "http://vocab.getty.edu/aat/300008389",
"type": "Type",
"_label": "City"
}
]
}
Types of Types
A common pattern is to not only classify main entities, but also to classify the types themselves in order to know what sort of type it is, without recognizing all of them individually. This is important when the set of first degree types is not easily enumerable, such as classifications for the type of object or work. If the set of vocabulary terms that can normally be used without further extension can be established, then this pattern is not used. This is also not needed when the relationship from the entity to the first degree type clarifies this, such as the motivation
or about
properties. This pattern is sometimes called a "meta-type" -- the type of a type.
Use cases for this pattern include:
- The classification is the type of object or work of the physical thing (e.g. Painting)
- The classification is the type of text of the statement (e.g. Dimensions Statement)
- The classification is the nationality of the person (e.g. Swiss)
The URIs for the second degree type (the "meta-type"), if present, MUST be drawn from the required vocabulary if it is listed there. For example, implementations MUST use the same AAT URI for Nationality as in the example below, however it is allowed to use this pattern with types that are not listed. It is encouraged to use appropriate AAT terms, and to bring the use cases to the community to discuss their inclusion in a future version. It is always possible to not list any meta-type, even for terms which happen to be covered by existing meta-types.
Example:
The object is classified as being a Painting, and the concept "Painting" is for classifying the type of object (as opposed to any other sort of classification).
{
"@context": "https://linked.art/ns/v1/linked-art.json",
"id": "https://linked.art/example/object/spring/3",
"type": "HumanMadeObject",
"_label": "Jeanne (Spring) by Manet",
"classified_as": [
{
"id": "http://vocab.getty.edu/aat/300033618",
"type": "Type",
"_label": "Painting",
"classified_as": [
{
"id": "http://vocab.getty.edu/aat/300435443",
"type": "Type",
"_label": "Type of Work"
}
]
},
{
"id": "http://vocab.getty.edu/aat/300133025",
"type": "Type",
"_label": "Artwork"
}
]
}
Rembrandt (a person) has a nationality of Dutch:
{
"@context": "https://linked.art/ns/v1/linked-art.json",
"id": "https://linked.art/example/person/rembrandt/1",
"type": "Person",
"_label": "Rembrandt",
"classified_as": [
{
"id": "http://vocab.getty.edu/aat/300111175",
"type": "Type",
"_label": "Dutch",
"classified_as": [
{
"id": "http://vocab.getty.edu/aat/300379842",
"type": "Type",
"_label": "Nationality"
}
]
}
]
}
Names and Identifiers for an Entity
Names
As the _label
property is intended only as internal documentation for the data itself, it is strongly RECOMMENDED that every entity that should be rendered to an end user also have at least one specific name. The name could be for an object, a person, a group, an event or anything else. This pattern uses the identified_by
property, with a Name
construct. The string value of the name is given in the content
property of the Name
.
It is somewhat unintuitive to think of a name as identifying the entity it is associated with, as names are typically not unique. However, as the name has its own class (Name) and could have an id
, it is not just an anonymous string and thus no longer a shared label. Instead each name is the particular instance of a name which is uniquely associated with its entity. With this formulation, the Name
instance does uniquely identify the entity.
If there is more than one name given, then there should be one that is classified_as
the primary name for use. This is done by adding the Primary Name (aat:300404670) term to it. There SHOULD be exactly one primary name given per language. Names are also part of human communication, and can have the Linguistic features of the model associated with them, such as having a particular language.
Example:
The primary name for "The Night Watch" is "The Night Watch" in English, and "De Nachtwacht" in Dutch:
{
"@context": "https://linked.art/ns/v1/linked-art.json",
"id": "https://linked.art/example/object/nightwatch/1",
"type": "HumanMadeObject",
"_label": "Night Watch by Rembrandt",
"identified_by": [
{
"type": "Name",
"classified_as": [
{
"id": "http://vocab.getty.edu/aat/300404670",
"type": "Type",
"_label": "Primary Name"
}
],
"content": "The Night Watch",
"language": [
{
"id": "http://vocab.getty.edu/aat/300388277",
"type": "Language",
"_label": "English"
}
]
},
{
"type": "Name",
"classified_as": [
{
"id": "http://vocab.getty.edu/aat/300404670",
"type": "Type",
"_label": "Primary Name"
}
],
"content": "De Nachtwacht",
"language": [
{
"id": "http://vocab.getty.edu/aat/300388256",
"type": "Language",
"_label": "Dutch"
}
]
}
]
}
Identifiers
Many entities are also given identifiers, such as accession numbers for objects, ORCIDs for people or groups, lot numbers for auctions, and so forth. Identifiers are represented in a very similar way to names, but instead use the Identifier
class. Identifiers will normally have a classification determining which sort of identifier it is, to distinguish between internal repository system assigned numbers from museum assigned accession numbers, for example.
As Identifier
s and Name
s use the same identified_by
property, the JSON will frequently have mixed classes in the array. Unlike Name
s, Identifier
s are not part of human language and thus cannot have a language associated with them.
Example:
The accession number identifier for "The Night Watch" is "SK-C-5":
{
"@context": "https://linked.art/ns/v1/linked-art.json",
"id": "https://linked.art/example/object/nightwatch/2",
"type": "HumanMadeObject",
"_label": "Night Watch by Rembrandt",
"identified_by": [
{
"type": "Identifier",
"classified_as": [
{
"id": "http://vocab.getty.edu/aat/300312355",
"type": "Type",
"_label": "Accession Number"
}
],
"content": "SK-C-5"
},
{
"type": "Name",
"classified_as": [
{
"id": "http://vocab.getty.edu/aat/300404670",
"type": "Type",
"_label": "Primary Name"
}
],
"content": "The Night Watch"
}
]
}
Equivalent Data URIs
Identifiers for the same entity within the web of linked data are treated differently from string identifiers such as accession numbers or ISBNs. In order to allow systems to follow the link and potentially process the information that they discover there, we use a new property called equivalent
to link out to other organizations' data about the same entity. For example, Rembrandt has a description in the Getty's ULAN vocabulary, in Wikidata, in the Library of Congress Name Authority File, and in OCLC's Virtual International Authority File, amongst many others. The URI given MUST identify the entity itself, rather than about page the entity. For example, in ULAN there is also the website version http://vocab.getty.edu/page/ulan/500011051 which MUST NOT be used with equivalent.
Note also that equivalent
can also be included when referencing entities across records in the API.
Example:
The Night Watch has an external URI that also identifies the same physical object in wikidata:
{
"@context": "https://linked.art/ns/v1/linked-art.json",
"id": "https://linked.art/example/object/nightwatch/3",
"type": "HumanMadeObject",
"_label": "Night Watch by Rembrandt",
"equivalent": [
{
"id": "https://www.wikidata.org/entity/Q219831",
"type": "HumanMadeObject",
"_label": "Night Watch"
}
]
}
Statements about an Entity
In many cases, current data does not support the level of specificity that the full ontology allows, or the information is simply best expressed in a human-readable form. For example, instead of a completely modeled set of parts with materials, many museum collection management systems allow only a single human-readable string for the "medium" or "materials statement". The same is true in many other situations, including rights or allowable usage statements, dimensions, edition statements and so forth. Any time that there is a description of the entity, with or without qualification as to the type of description, then this pattern can be used to record the descriptive text.
The pattern makes use of the LinguisticObject
class that is used to identify a particular piece of textual content. These Linguistic Objects are then referred to by any other entity. They maintain the statement's text in the content
property, and the language of the statement (if known) in the language
property. If the content is not plain text, then using format
to give the media type is recommended (e.g. "text/html" for HTML)
Note that, in the Linked Art API, the reference to a Language from a Statement or Name can also have a notation
property which gives the commonly used two or three letter code for it, if known.
Use cases for this pattern include:
- General description of the entity
- Materials statement for an object
- Attribution statement for an image
- Biography for a person
- Dimensions statement for a part of an object
Note that both Names and Statements can have their own Names (e.g. "Former Title", "Supplied Description") and their own Statements (such as the source of the name or statement), however best practice is to only use these sparingly.
Example:
Having only a textual description of the materials in English, the content "Oil on Canvas"
is recorded as referring to the painting as a "materials" (aat:300435429) statement:
{
"@context": "https://linked.art/ns/v1/linked-art.json",
"id": "https://linked.art/example/object/nightwatch/4",
"type": "HumanMadeObject",
"_label": "Night Watch by Rembrandt",
"referred_to_by": [
{
"type": "LinguisticObject",
"classified_as": [
{
"id": "http://vocab.getty.edu/aat/300435429",
"type": "Type",
"_label": "Material Statement",
"classified_as": [
{
"id": "http://vocab.getty.edu/aat/300418049",
"type": "Type",
"_label": "Brief Text"
}
]
}
],
"content": "Oil on Canvas",
"language": [
{
"id": "http://vocab.getty.edu/aat/300388277",
"type": "Language",
"_label": "English"
}
]
}
]
}
Events and Activities
Linked Art frequently uses an intermediate entity that represents an activity to connect between other entities. For example instead of having a direct relationship between the artwork and the artist, there is an activity that represents creating the artwork, which is then carried out by the artist, but can also have many other properties. This enables us to associate time, place, actors, techniques, other objects and more with the creation activity, rather than needing many individual relationships. Combined with the classification pattern, it allows any sort of activity to be modeled without having to create many specific relationships for each case.
The key participants in those different types of events are:
- the Object being acted upon, or used to perform the activity,
- the Actors (people or organizations) that perform the activity,
- the Locations at which the activity occurs,
- the Time at which it occurs, and/or the Period during which it occurs,
- any other Event or activity which caused it,
- and a classification for the technique used.
These can be thought of as What, Who, Where, When, Why and How respectively.
The general pattern is to create data for the event (with the class Event
) or activity (with the class Activity
or a more specific class) within the "main" record for the entity that is created, destroyed or primarily affected by the event, and then to associate the participants with that embedded data construct. The relationships for time (timespan
and/or during
) and place (took_place_at
) are relevant to Event
s that happen without the direct cause being a human action, and the relationship for the actor (carried_out_by
) is added to those for human activities. The relationship to the object is dependent on the type of event or activity, which are discussed in more detail in the specific sections.
There are both subclasses, such as Birth
, Production
and Creation
, and classifications associated with them to be more specific, such as glassblowing (aat:300053932) to clarify the type or technique of the activity. There are three common categories of activity which are used across the different entity types: their beginning of existence, their end of existence, and core activities that they either performed (for people or groups) or were required for (for objects and works). The table below summaries the beginning and ending of existence classes per main entity class. Note that conceptual entities cannot have an end of existence, and Places have neither.
Class | Beginning | Ending |
---|---|---|
HumanMadeObject |
Production |
Destruction |
DigitalObject |
Creation |
Erasure |
LinguisticObject |
Creation |
None |
VisualItem |
Creation |
None |
PropositionalObject |
Creation |
None |
Type (& below) |
Creation |
None |
Set |
Creation |
None |
Person |
Birth |
Death |
Group |
Formation |
Dissolution |
Example:
"Spring" was produced in a Production activity carried out by Manet in 1881, somewhere in France.
{
"@context": "https://linked.art/ns/v1/linked-art.json",
"id": "https://linked.art/example/object/spring/4",
"type": "HumanMadeObject",
"_label": "Jeanne (Spring) by Manet",
"produced_by": {
"type": "Production",
"timespan": {
"type": "TimeSpan",
"_label": "1881",
"begin_of_the_begin": "1881-01-01T00:00:00Z",
"end_of_the_end": "1881-12-31T23:59:59Z"
},
"took_place_at": [
{
"id": "https://linked.art/example/place/france",
"type": "Place",
"_label": "France"
}
],
"carried_out_by": [
{
"id": "https://linked.art/example/person/manet",
"type": "Person",
"_label": "Manet"
}
],
"during": [
{
"id": "https://linked.art/example/event/19c",
"type": "Period",
"_label": "19th Century"
}
]
}
}
Time Span Details
The minimal timespan model is given above, with just the begin_of_the_begin
and an end_of_the_end
properties to record the beginning of the span, and the end of the span. The end of the span is included in the span, and thus if the end_of_the_end
is "1500-01-01T00:00:00Z"
, then the very first second of 1500 is included, and thus the timestamp should be "1499-12-31T23:59:59Z" for the last moment of the 1400s and not any of 1500. (This is a requirement from the underlying CIDOC-CRM ontology)
The datestamps must be fully qualified with year, month, day, hour, minute and second. If a timezone is not supplied it should be assumed to be "Z" (GMT) to ensure comparisons work correctly. This is the "xsd:dateTime" profile of ISO8601.
It is very useful to have a Name
for the TimeSpan
that gives a human readable version of the machine readable timestamps and durations. This uses the pattern given above for naming things, and might be displayed to user directly, whereas the other properties could be used for matching a time-based search.
There are other properties for TimeSpan
instances that are useful when the exact span of time is not certain.
end_of_the_begin
: This property is another datetime that gives the latest possible datetime for the beginning of the span.begin_of_the_end
: This property is yet another datetime that gives the earliest possible datetime for the end of the span.duration
: This property refers to aDimension
, that describes the amount of time that the activity or event took, within the span of time. As the property defines the type of dimension, theDimension
instance does not need to beclassified_as
any particularType
... it is always a duration.
Example:
The Christie's auction of the Stowe House took place over 40 days in August and September, 1848 (Source).
{
"@context": "https://linked.art/ns/v1/linked-art.json",
"id": "https://linked.art/example/event/stowe/1",
"type": "Activity",
"_label": "Auction of Stowe House",
"classified_as": [
{
"id": "http://vocab.getty.edu/aat/300054751",
"type": "Type",
"_label": "Auction Event"
}
],
"timespan": {
"type": "TimeSpan",
"identified_by": [
{
"type": "Name",
"content": "40 days in August and September, 1848"
}
],
"begin_of_the_begin": "1848-08-01T00:00:00Z",
"end_of_the_begin": "1848-08-20T23:59:59Z",
"begin_of_the_end": "1848-09-09T00:00:00Z",
"end_of_the_end": "1848-09-30T23:59:59Z",
"duration": {
"type": "Dimension",
"value": 3,
"unit": {
"id": "http://vocab.getty.edu/aat/300379242",
"type": "MeasurementUnit",
"_label": "days"
}
}
}
}
Parts
Describing a hierarchy of parts of entities is a core pattern for having increasingly granular or specific descriptions. The advantage of partitioning is that more specific information can be provided about each part, as an aspect separate from the whole. This pattern covers the spectrum of different classes used in the model, from physical and textual, to temporal or geographic. Parts are given using the properties part_of
, from the part to the whole.
Use cases for this pattern include:
- Delineating physical components or aspects of an object, such as the frame being part of the painting, or the recto side being part of the page. Physical parts may be separable from the object (the frame) or not (the recto of a single page) -- they are both described using the same pattern.
- Separating a long period or event into smaller sections, such as battles within a war, or the activities of different people during the production of an object.
- Separating geographical locations into smaller subdivisions, such as that the neighborhood is part of the city, which is in turn part of a county or district.
- Partitioning a text into segments, such as that the paragraph is part of a chapter, which is in turn part of the entire book
Note that parts are generally separate records from the main entity, and the main entity does not refer to its parts. Instead the API defines a method to retrieve all of the parts of the main entity, or other incoming references.
Example:
The canvas support is part of the overall painting of Spring.
{
"@context": "https://linked.art/ns/v1/linked-art.json",
"id": "https://linked.art/example/object/spring/support",
"type": "HumanMadeObject",
"_label": "Support of Spring",
"classified_as": [
{
"id": "http://vocab.getty.edu/aat/300014844",
"type": "Type",
"_label": "Support",
"classified_as": [
{
"id": "http://vocab.getty.edu/aat/300241583",
"type": "Type",
"_label": "Part Type"
}
]
}
],
"referred_to_by": [
{
"type": "LinguisticObject",
"classified_as": [
{
"id": "http://vocab.getty.edu/aat/300435429",
"type": "Type",
"_label": "Material Statement",
"classified_as": [
{
"id": "http://vocab.getty.edu/aat/300418049",
"type": "Type",
"_label": "Brief Text"
}
]
}
],
"content": "Canvas"
}
],
"part_of": [
{
"id": "https://linked.art/example/object/spring",
"type": "HumanMadeObject",
"_label": "Jeanne (Spring) by Manet"
}
]
}
Membership
Membership in a set, collection or group is treated slightly differently. A set can have no members and still be a set, whereas if you destroy the last part of an object, it no longer exists. For example, a department of an organization (a Group) might not have any members due to the retirement of the last member, but there is a still an identifiable, ongoing group that would hopefully gain members when new hires are made. Membership is given using the member_of
property, instead of the corresponding part_of
.
Use cases for the membership pattern include:
- A curator is a member of the paintings department of the museum.
- A painting is part of the collection of a the museum, or of the collection managed by the paintings department.
Example:
Rembrandt was a member of the Guild of St Luke (a guild of painters).
{
"@context": "https://linked.art/ns/v1/linked-art.json",
"id": "https://linked.art/example/person/rembrandt/2",
"type": "Person",
"_label": "Rembrandt",
"member_of": [
{
"id": "https://linked.art/example/group/stluke",
"type": "Group",
"_label": "Guild of St Luke"
}
]
}
Partitioning of Internal Constructs
It is sometimes necessary to partition an activity that is embedded within another record, such as the Production
of an object or Creation
of a work, in order to be more specific about individual roles or aspects. In this case, as there isn't another record to refer to with part_of
, the parts are included within the record using part
. This allows information to be associated with either the individual parts of the activity, or the whole activity.
Example:
A video (currently modeled as a Work that contains language, or a LinguisticObject
) about Rembrandt's Night Watch was directed by Peter Greenaway and Produced by Femke Wolting.
{
"@context": "https://linked.art/ns/v1/linked-art.json",
"id": "https://linked.art/example/text/rembrandtjaccuse/1",
"type": "LinguisticObject",
"_label": "Rembrandt's J'accuse",
"created_by": {
"type": "Creation",
"timespan": {
"type": "TimeSpan",
"_label": "2008",
"begin_of_the_begin": "2008-01-01T00:00:00Z",
"end_of_the_end": "2008-12-31T23:59:59Z"
},
"part": [
{
"type": "Creation",
"classified_as": [
{
"id": "http://vocab.getty.edu/aat/300025654",
"type": "Type",
"_label": "Director"
}
],
"carried_out_by": [
{
"id": "https://linked.art/example/person/greenaway",
"type": "Person",
"_label": "Peter Greenaway"
}
]
},
{
"type": "Creation",
"classified_as": [
{
"id": "http://vocab.getty.edu/aat/300197742",
"type": "Type",
"_label": "Producer"
}
],
"carried_out_by": [
{
"id": "https://linked.art/example/person/wolting",
"type": "Person",
"_label": "Femke Wolting"
}
]
}
]
}
}