wikidataloader

Easy pythonic wrapper around the Wikidata SPARQL API for quick creation of datasets from Wikidata.

Only supports simple, non-recursive queries - for complex queries please directly use the SPARQL API provided by Wikidata.

It does not support complex operators (ordering, datetime conversion, string/numeric filtering etc.), because these can be substituted by preprocessing the dataset in Python after retrieval.

Usage

Look up the URIs for properties (e.g. P31) and objects (e.g. Q5) on Wikidata's search engine.

from wikidataloader import WikidataQuery

# Linguists from Germany with birth places and gender

results = WikidataQuery.search(
    # {is_instance:human, country_of_origin:Germany, profession:linguist}
    filters={"P31": "Q5", "P27": "Q183", "P106": "Q14467526"}, 

    # selects the properties "Gender" and "Birth Place" as columns in the dataframe and names them "Gender" and "City of Birth"
    select=[("P21", "Gender"), ("P19", "City of Birth")],

    # returns a maximum of 5 results
    limit=5,

    # retrieves labels in English, if available
    default_language="en" 
).to_pandas()

results

>>>                                    item Gender        City_of_Birth
>>> 0                        Hermann Weller   male     Schwäbisch Gmünd
>>> 1                             Hans Wehr   male              Leipzig
>>> 2                       Theodor Haecker   male            Mulfingen
>>> 3                   Gottfried Bernhardy   male  Gorzów Wielkopolski
>>> 4                    Wilhelm Streitberg   male   Rüdesheim am Rhein

For more examples, see example.ipynb

Install

Install using pip:

pip install wikidataloader

Limitations

• Does not support recursive queries
• Does not support Senses and Forms for Lexeme queries