User:Push-f/Wikidata:Limitations

This page attempts to document the limitations of Wikidata.

Limitations imposed by policies

License

All data in Wikidata must comply with Wikidata:Copyright, which requires freedom from copyright restrictions.

Privacy

Statements about living people are subject to the Wikidata:Living people policy (which isn't summarized here because it's a bit unclear).

Notability

Wikidata has a Notability policy. Items should only be created if they meet the criteria outlined in that policy. Items that do not meet the policy are regularly deleted.

Verifiability

The majority of statements on Wikidata should be verifiable: they should be supported by a source of information such as a book, scientific publication, or newspaper article. In Wikidata, references are used to point to specific sources that back up the data provided in a statement.

Permanent limitations

The following limitations are inherent to the mission and nature of Wikidata and will likely stay with Wikidata forever.

Accuracy

Sometimes statements on Wikidata are wrong without having a deprecated rank. The main causes for such wrong statements are:

• accidentally using the wrong item or property
  • may happen when you only read the label but disregard the description
  • may happen when the label/description are confusing, misleading or out-of-date
  • restrictive qualifiers may be mixed up with non-restrictive qualifiers because the Wikibase UI doesn't indicate the difference between them at all (which is also difficult because some qualifiers can be both)
• conflation (when one item represents different things), for example:
  • One item might have sitelinks to Wikipedia articles describing different concepts, which can lead to Wikidata contributors that speak different languages to disagree about what the item represents.
  • When splitting up a conflated item you might forget to update the backlinks accordingly.
• transitivity — instance of (P31) or subclass of (P279) statements are often added without considering all of the implications (the Ontology WikiProject works on tracking down such wrong implications), as result it is impossible to reliably check whether given Wikidata item describes an event or a physical object
• diverging labels/descriptions between languages — if a label or description in one language is changed other translators will generally never know about it
• vandalism (when people deliberately attempt to damage or compromise the integrity of Wikidata)
• common misconceptions

Coverage

Main page: Wikidata:Coverage

No knowledge base can have an entry for everything because some classes have infinitely many instances e.g. there are infinitely many numbers. Because of the notability policy mentioned in the previous section Wikidata does not and will not ever have an item for every instance of many classes because they don't meet the notability criteria. However just because something is notable does not mean that Wikidata has an item about it: Wikidata is lacking millions of items about notable things or concepts. Thousands of items are being created every single day.

API & data format stability

The API and data formats used by Wikidata are subject to change as per the Stable Interface Policy.

Current limitations

The following limitations are limitations to the status quo and may be resolved in the future.

User interface

Main pages: /User interface & /WDQS

The Wikidata user interface has several limitations and many tools have been developed to address them. The development of the Wikidata user interface is very much still ongoing.

Data model

• Wikibase has no first-level support for property scopes/constraints, these are instead implemented via an extension that only poorly integrates with Wikibase (violations are only reported after they have been made, absolutely no integration with wbeditentity, etc.)
• Wikibase has no first-level support for the distinction of whether or not a qualifier property is restrictive or non-restrictive (which would be really important so that data contributors and consumers correctly interpret the statements).
• Wikibase has no first-level support for negated statements.
• Wikibase stores values in a specific order, which is confusing because it might mislead people into thinking that the storage order matters, which it does not and inconvenient because API consumers need to reorder values to reliably get a semantic order. While semantic order can be expressed via qualifiers, like series ordinal (P1545) and publication date (P577), these qualifiers are not taken into account by the UI or the API when they list values. phab:T173432

Data types

Wikidata currently supports 17 data types, several of these have limitations:

• Several string-based data types have a length limit as defined with wmgWikibaseStringLimits in InitialiseSettings.php. These limits do not appear to be discoverable via the API (other than running into them). The current work around appears to be to split up the value across multiple claims and to qualify them with nature of statement (P5102)value split up, due to exceeding the 1,500 character limit on strings (Q110418529)series ordinal (P1545)X where X is the index of the part.
• String doesn't support horizontal tabs or vertical space characters. phab:T323705
• String cannot begin or end with whitespace.
• Point in time doesn't support time of day. phab:T57755
• Quantity doesn't support infinity (Q205), e.g. you cannot state prime number (Q49008)quantity (P1114)infinitely many. phab:T324403

Such issues are tracked on Phabricator, which is also where the introduction of new data types can be proposed.

API limitations

• the data types supported by a Wikibase instance currently cannot be listed via the MediaWiki action API or the Wikibase Lua API
• the Wikibase system ontology (Q115686989) is currently apparently not automatically generated by Wikibase but instead hardcoded and even misses several definitions
• the serialization library puts data in an inefficient order phab:T323618

Linking non-free media

Wikidata only support one data type that is specifically for media, which is Commons media file. Wikimedia Commons however only supports media that is available under a free license.

This means for example that Homer Simpson (Q7810) has no image (P18) property because there is no direct image of him available under a free license.

The English Wikipedia does have an image of him w:File:Homer Simpson 2006.png, however it cannot be linked via image (P18).

There do currently two subproperties of image (P18) of datatype URL that allow the linking of non-free media:

• non-free artwork image URL (P6500)
• podcast logo URL (P10286)

however since these only have the datatype URL data consumers cannot easily recognize that these point to media files (and the Wikidata user interface also doesn't display files linked via these properties).

Linking non-Wikimedia wikis

Sitelinks only work for Wikimedia sites and cannot be used to link other non-Wikimedia wikis.

For example Template:Wikidata entity link (Q17280715) cannot be linked to osmwiki:Template:Q.

Sometimes properties are created to still be able to link other wikis via formatter URL (P1630) e.g. OpenStreetMap tag or key (P1282).

See also Wikidata property linking to external MediaWiki wiki (Q62619638) and a list of such instances.