Wikidata

Main parts

Fundamentally, an item consists of:

• An identifier (the QID), related to a label and a description.
• Optionally, multiple aliases and some number of statements (and their properties and values).

Statements

Statements are how any information known about an item is recorded in Wikidata. Formally, they consist of key–value pairs, which match a property (such as "author", or "publication date") with one or more entity values (such as "Sir Arthur Conan Doyle" or "1902"). For example, the informal English statement "milk is white" would be encoded by a statement pairing the property color (P462) with the value white (Q23444) under the item milk (Q8495).

Statements may map a property to more than one value. For example, the "occupation" property for Marie Curie could be linked with the values "physicist" and "chemist", to reflect the fact that she engaged in both occupations.^[6]

Values may take on many types, including other Wikidata items, strings, numbers, or media files. Properties prescribe what types of values they may be paired with. For example, the property official website (P856) may only be paired with values of type "URL".^[7]

Optionally, qualifiers can be used to refine the meaning of a statement by providing additional information. For example, a "population" statement could be modified with a qualifier such as "point in time (P585): 2011" (as its own key-value pair). Values in the statements may also be annotated with references, pointing to a source backing up the statement's content.^[8] As with statements, all qualifiers and references are property–value pairs.

Properties

Each property has a numeric identifier prefixed with a capital P and a page on Wikidata with an optional label, description, aliases, and statements. As such, there are properties with the sole purpose of describing other properties, such as subproperty of (P1647).

Properties may also define more complex rules about their intended usage, termed constraints. For example, the capital (P36) property includes a "single value constraint", reflecting the reality that (typically) territories have only one capital city. Constraints are treated as testing alerts and hints, rather than inviolable rules.^[9]

Before a new property is created, it needs to undergo a discussion process.^[10][11]

The most used property is cites work (P2860), which is used on more than 290,000,000 item pages as of November 2023.^[update][12]

Lexemes

In linguistics, a lexeme is a unit of lexical meaning representing a group of words that share the same core meaning and grammatical characteristics.^[13][14] Similarly, Wikidata's lexemes are items with a structure that makes them more suitable for storing lexicographical data. Since 2016, Wikidata has supported lexicographical entries in the form of lexemes.^[15]

In Wikidata, lexicographical entries have a different identifier from regular item entries. These entries are prefixed with the letter L, such as in the example entries for book (L16168) and cow (L3726). Lexicographical entries in Wikidata can contain statements, senses, and forms.^[16] The use of lexicographical entries in Wikidata allows for the documentation of word usage, the connection between words and items on Wikidata, word translations, and enables machine-readable lexicographical data.

In 2020, lexicographical entries on Wikidata exceeded 250,000. The language with the most lexicographical entries was Russian, with a total of 101,137 lexemes, followed by English with 38,122 lexemes. There are over 668 languages with lexicographical entries on Wikidata.^[17]

Entity schemas

In Wikidata, a schema is a data model that outlines the necessary attributes for a data item.^[18][19] For instance, a data item that uses the attribute "instance of" with the value "human" would typically include attributes such as "place of birth," "date of birth," "date of death," and "place of death."^[20] The entity schema in Wikidata utilizes Shape Expression (ShEx) to describe the data in Wikidata items in the form of a Resource Description Framework (RDF).^[21] The use of entity schemas in Wikidata helps address data inconsistencies and unchecked vandalism.^[18]

In January 2019, development started of a new extension for MediaWiki to enable storing ShEx in a separate namespace.^[22][23] Entity schemas are stored with different identifiers than those used for items, properties, and lexemes. Entity schemas are stored with an "E" identifier, such as E10 for the entity schema of human data instances and E270 for the entity schema of building data instances. This extension has since been installed on Wikidata^[24] and enables contributors to use ShEx for validating and describing Resource Description Framework data in items and lexemes. Any item or lexeme on Wikidata can be validated against an entity schema,^{[clarification needed]} and this makes it an important tool for quality assurance.

Initial rollout

Wikidata was launched on 29 October 2012 and was the first new project of the Wikimedia Foundation since 2006.^[2][33][34] At release, only the centralization of language links was available. This enabled items to be created and filled with basic information: a label – a name or title, aliases – alternative terms for the label, a description, and links to articles about the topic in all the various language editions of Wikipedia (interwikipedia links).

Historically, a Wikipedia article would include a list of interlanguage links (links to articles on the same topic in other editions of Wikipedia, if they existed). Wikidata was originally a self-contained repository of interlanguage links.^[35] Wikipedia language editions were still not able to access Wikidata, so they needed to continue to maintain their own lists of interlanguage links.^{[citation needed]}

On 14 January 2013, the Hungarian Wikipedia became the first to enable the provision of interlanguage links via Wikidata.^[36] This functionality was extended to the Hebrew and Italian Wikipedias on 30 January, to the English Wikipedia on 13 February and to all other Wikipedias on 6 March.^{[37][38][39][40]} After no consensus was reached over a proposal to restrict the removal of language links from the English Wikipedia,^[41] they were automatically removed by bots. On 23 September 2013, interlanguage links went live on Wikimedia Commons.^[42]

Statements and data access

On 4 February 2013, statements were introduced to Wikidata entries. The possible values for properties were initially limited to two data types (items and images on Wikimedia Commons), with more data types (such as coordinates and dates) to follow later. The first new type, string, was deployed on 6 March.^[43]

The ability for the various language editions of Wikipedia to access data from Wikidata was rolled out progressively between 27 March and 25 April 2013.^[44][45] On 16 September 2015, Wikidata began allowing so-called arbitrary access, or access from a given article of a Wikipedia to the statements on Wikidata items not directly connected to it. For example, it became possible to read data about Germany from the Berlin article, which was not feasible before.^[46] On 27 April 2016, arbitrary access was activated on Wikimedia Commons.^[47]

According to a 2020 study, a large proportion of the data on Wikidata consists of entries imported en masse from other databases by Internet bots, which helps to "break down the walls" of data silos.^[48]

Query service and other improvements

On 7 September 2015, the Wikimedia Foundation announced the release of the Wikidata Query Service,^[49] which lets users run queries on the data contained in Wikidata.^[50] The service uses SPARQL as the query language. As of November 2018, there are at least 26 different tools that allow querying the data in different ways.^[51] It uses Blazegraph as its triplestore and graph database.^[52][53]

In 2021, Wikimedia Deutschland released the Query Builder,^[54] "a form-based query builder to allow people who don't know how to use SPARQL" to write a query.

The Wikidata Embedding Project was made available in October 2025. It provides a vector-based semantic search tool, allowing plain-language queries, and supports the Model Context Protocol standard that makes the data more readily available to AI systems.^[55] The project is a partnership between Wikimedia Deutschland, Jina.AI and DataStax, an IBM subsidiary.^[55]

Logo

The bars on the logo contain the word "WIKI" encoded in Morse code.^[56] It was created by Arun Ganesh and selected through community decision.^[57]