node package manager
Loosely couple your services. Use Orgs to version and reuse your code. Create a free org »



Command-line tool to extract taxonomies from Wikidata.


wikidata-taxonomy requires at least NodeJs version 6.

Install globally to make command wdtaxonomy accessible from your shell $PATH:

$ npm install -g wikidata-taxonomy


This module provides the command wdtaxonomy. By default, a usage help is printed:

$ wdtaxonomy

  Usage: wdtaxonomy [options] <id>

  extract taxonomies from Wikidata


    -V, --version                        output the version number
    -b, --brief                          omit counting instances and sites
    -c, --children                       get direct subclasses only
    -C, --color                          enforce color output
    -d, --descr                          include item descriptions
    -e, --sparql-endpoint <url>          customize the SPARQL endpoint
    -f, --format <tree|csv|json|ndjson>  output format
    -i, --instances                      include instances
    -I, --no-instancecount               omit counting instances
    -j, --json                           use JSON output format
    -l, --lang <lang>                    specify the language to use
    -L, --no-labels                      omit all labels
    -m, --mappings <ids>                 mapping properties (e.g. P1709)
    -n, --no-colors                      disable color output
    -o, --output <file>                  write result to a file
    -P, --property <id>                  hierarchy property (e.g. P279)
    -p, --post                           use HTTP POST to disable caching
    -r, --reverse                        get superclasses instead
    -s, --sparql                         print SPARQL query and exit
    -S, --no-sitecount                   omit counting sites
    -t, --total                          count total number of instances
    -u, --user <name>                    user to the SPARQL endpoint
    -U, --uris                           show full URIs in output formats
    -v, --verbose                        make the output more verbose
    -w, --password <string>              password to the SPARQL endpoint
    -h, --help                           output usage information

The first arguments needs to be a Wikidata identifier to be used as root of the taxonomy. For instance extract a taxonomy of planets (Q634):

$ wdtaxonomy Q634

To look up by label, use wikidata-cli.

The extracted taxonomy by default is based on statements using the property "subclass of" (P279) or "subproperty of" (P1647). Taxonomy extraction and output can be controlled by several options. Option --sparql (or -s) prints the underlying SPARQL queries instead of executing them.


Direct subclasses of planet (Q634) with description and mappings:

$ wdtaxonomy.js Q634 -c -d -m =

The hierarchy properties P279 ("subclass of") and P31 ("instance of") to build taxonomies from can be changed with option property (-P).

Members of (P463) the European Union (Q458):

$ wdtaxonomy Q458 -P P463

Members of (P463) the European Union (Q458) and number of its citizens in Wikidata (P27):

$ wdtaxonomy Q458 -P 463/27

Wikiversity (Q370) editions mapped to their homepage URL (P856):

$ wdtaxonomy Q370 -i -m P856

Biological taxonomy of mammals (Q7377):

$ wdtaxonomy Q7377 -P P171 --brief

Property constraints (Q21502402) with number of properties that have each constraint:

$ wdtaxonomy Q21502402 -P 279,2302 

As Wikidata is no strict ontology, subproperties are not factored in. For instance this query does not include members of the European Union although P463 is a subproperty of P361.

Parts of (P361) the European Union (Q458):

$ wdtaxonomy Q458 -P P361

A taxonomy of subproperties can be queried like taxonomies of items. The hierarchy property is set to P1647 ("subproperty of") by default:

$ wdtaxonomy P361
$ wdtaxonomy P361 -P P1647  # equivalent

Subproperties of "part of" (P361) and which of them have an inverse property (P1696):

$ wdtaxonomy P361 -P P1647/P1696

Inverse properties are neither factored in so queries like these do not necesarrily return the same results:

What hand (Q33767) is part of (P361):

$ wdtaxonomy Q33767 -P 361 -r

What parts the hand (Q33767) has (P527):

$ wdtaxonomy Q33767 -P 527


Query options

brief (-b)

Don't count instance and sites. Same as -S/--no-sitecount and -I/--no-instancecount.

children (-c)

Get direct subclasses only

descr (-d)

Include item descriptions

sparql-endpoint (-e)

SPARQL endpoint to query (default:

instances (-i)

Include instances

no-instancecount (-I)

Don't count number of instances

lang (-l)

Language to get labels in (default: en)

no-labels (-L)

Omit all labels. This allows for querying larger taxonomies (several thousands of classes), especially if combined with option --brief.

mappings (-m)

Lookup mappings based on given comma-separated properties such as P1709. The following keywords can be used as shortcuts:

  • equal or =: equivalent property (P1628), equivalent class (P1709), and exact match (P2888)
  • broader: external superproperty (P2235)
  • narrower: narrower external class (P3950), external subproperty (P2236)
  • class: properties for mapping classes
  • property: properties for mapping properties
  • all all properties for ontology mapping (instances of Q30249126)

reverse (-r)

Get superclasses instead of subclasses up to the root

no-sitecount (-I)

Don't count number of sites

total (-t)

Count total (transitive) number of instances, including instances of subclasses

post (-p)

Use HTTP POST to disable caching

sparql (-s)

Don't actually perform a query but print SPARQL query and exit

user (-u)

User to the SPARQL endpoint

password (-w)

Password to the SPARQL endpoint

Output options

color (-C)

enable color output if it's disabled (e.g. when output is piped or written to a file)

format (-f)

Output format

json (-j)

Use JSON output format. Same as --format json but shorter.

no-colors (-n)

disable color output

output (-o)

write result to a file given by name

uris (-U)

Show full URIs in output formats, e.g. instead of Q1

verbose (-v)

Show verbose error messages

Output formats

Tree format

By default, the taxonomy is printed in "tree" format with colored Unicode characters:

$ wdtaxonomy Q17362350
planet of the Solar System (Q17362350) •2 ↑
├──outer planet (Q30014) •25 ×4 ↑↑
└──inner planets (Q3504248) •8 ×4 ↑↑

The output contains item labels, Wikidata identifiers, the number of Wikimedia sites connected to each item (indicated by bullet character ""), the number of instances (property P31), indicated by a multiplication sign "×"), and an upwards arrow ("") as indicator for additional superclasses.

Option "--instances" (or "-i") explicitly includes instances:

$ wdtaxonomy -i Q17362350
planet of the Solar System (Q17362350) •2 ↑
├──outer planet (Q30014) •25 ↑↑
|   -Saturn (Q193)
|   -Jupiter (Q319)
|   -Uranus (Q324)
|   -Neptune (Q332)
└──inner planets (Q3504248) •8 ↑↑
    -Earth (Q2)
    -Mars (Q111)
    -Mercury (Q308)
    -Venus (Q313)

Classes that occur at multiple places in the taxonomy (multihierarchy) are marked like in the following example:

$ wdtaxonomy Q634
planet (Q634) •202 ×7 ↑
├──extrasolar planet (Q44559) •88 ×833 ↑
|  ├──circumbinary planet (Q205901) •15 ×10
|  ├──super-Earth (Q327757) •32 ×46 ↑
├──terrestrial planet (Q128207) •70 ×7
|  ╞══super-Earth (Q327757) •32 ×46 ↑ …

JSON format

Option --format json serializes the taxonomy as JSON object. The format follows specification of JSKOS Concept Schemes:

  "type": [ "" ],
  "modified": "2017-11-06T10:25:54.966Z",
  "license": [
      "uri": "",
      "notation": [ "CC0" ]
  "languages": [ "en" ],
  "topConcepts": [
    { "uri": "" }
  "concepts": [ ]

Field concepts contains an array of all extracted Wikidata entities (usually classes and instances) as JSKOS Concepts:

  "uri": "",
  "notation": [ "Q17362350" ],
  "prefLabel": {
    "en": "planet of the Solar System"
  "scopeNote": {
    "en": [ "inner and outer planets of our solar system" ]
  "broader": [
    { "uri": "" }
  "narrower": [
    { "uri": "" },
    { "uri": "" }

Instances (option --instances) are linked via field subjectOf the same way as field broader and narrower.

The number of instances and sites, if counted is given as array of [JSKOS Concept Occurrences](( in field occurrences, each identified by subfield relation:

  "uri": "",
  "notation": [ "Q30014" ],
  "prefLabel": {
    "en": "outer planet of the Solar system"
  "occurrences": [
      "relation": "",
      "count": 4
      "relation": "",
      "count": 25

Mappings (option --mappings) are stored in field mappings as array of JSKOS Concept Mappings:

    "from": {
      "memberSet": [
        { "uri": "" }
    "to": {
      "memberSet": [
        { "uri": "" }
    "type": [

The mapping type is given in field type with the Wikidata property URI as last array element and the SKOS mapping relation URI as first.

NDJSON format

Option --format ndjson serializes JSON field concepts with one record per line. The order if records is same as in tree, json, and csv format but each concept is only included once.

CSV format

The CSV format ("--format csv") is optimized for comparing differences in time. Each output row consists of five fields:

  • level in the hierarchy indicated by zero or more "-" (default) or "=" characters (multihierarchy).

  • id of the item. Items on the same level are sorted by their id.

  • label of the item. Language can be selected with option --language. The character , in labels is replaces by a whitespace.

  • sites: number of connected sites (Wikipedia and related project editions). Larger numbers may indicate more established concepts.

  • parents outside of the hierarchy, indicated by zero or more "^" characters.

For instance the CSV output for Q634 would be like this:

$ wdtaxonomy -f csv Q634
-,Q44559,extrasolar planet,81,833,^
--,Q205901,circumbinary planet,14,10,
-,Q128207,terrestrial planet,67,7,

In this example there are 196 Wikipedia editions or other sites with an article about planets and seven Wikidata items are direct instance of a planet. At the end of the line "^" indicates that "planet" has one superclass. In the next rows "extrasolar planet" (Q44559) is a subclass of planet with another superclass indicated by "^". Both "circumbinary planet" and "super-Earth" are subclasses of "extrasolar planet". The latter also occurs as subclass of "terrestrial planet" where it is marked by "==" instead of "--".

Usage as module

const { queryTaxonomy, serializeTaxonomy } = require('wikidata-taxonomy')
// serialize taxonomy to stream
serializeTaxonomy.csv(taxonomy, process.stdout)
serializeTaxonomy.text(taxonomy, process.stdout, {colors: true})
serializeTaxonomy.json(taxonomy, process.stdout)
serializeTaxonomy.ndjson(taxonomy, process.stdout)

Usage in web applications

Requires wikidata-sdk and jQuery (but no other dependencies).

See directory example for a demo.

Release notes

Release notes are listed in file in the source code repository.

See Also

build status npm version Documentation Status

This document

Related tools