wikimedica-disease-search/data/wikimedica/alignement.py

import requests
import rdflib

SIGNES_ET_SYMPTOMES_WIKIMEDICA_PAGE = (
    'http://wikimedi.ca/wiki/Sp%C3%A9cial:'
    'Export_RDF/Concept:Signes_et_sympt%C3%B4mes'
)

PREFIX_PROPERTY = "http://wikimedi.ca/wiki/Special:URIResolver/Property-3A"

WIKIDATA_SPARQL_ENDPOINT = "https://query.wikidata.org/sparql"

USER_AGENT = "WikimedicaDiseaseSearch/0.1 (https://gitlab.com/matteodelabre/wikimedica-disease-search)"

session = requests.Session()
session.headers = {'User-Agent': USER_AGENT}

def make_sparql_request_wikimedica(page, request):
    """
    Effectue une requête SPARQL sur une page de WikiMedica

    :param page: Page de WikiMedica ciblé
    :param request: Requête SPARQL appliqué
    """
    data = session.get(page, stream=True)

    g = rdflib.Graph()
    g.parse(data.raw)
    qres = g.query(request)

    return [row.asdict() for row in qres]


def make_sparql_request_endpoint(endpoint, request):
    """
    Effectue une requête SPARQL sur point d'accés

    :param page: Adresse du point d'accés
    :param request: Requête SPARQL appliqué
    """
    res = session.get(endpoint, params={'format': 'json' ,'query': request})

    if res.status_code != 200:
        raise Exception('Erreur {}'.format(res.status_code))

    return res.json()


nb_wikimedica_entity = 0
nb_align_by_property = 0
nb_align_by_name = 0
nb_align_conflict = 0

# get name and URI from WikiMedica
request_name_uri = """
    SELECT ?name ?uri
       WHERE {
          ?el rdfs:label ?name .
          ?el rdfs:isDefinedBy ?uri
       }
"""

entitys_dic = make_sparql_request_wikimedica(
    SIGNES_ET_SYMPTOMES_WIKIMEDICA_PAGE,
    request_name_uri
)

nb_wikimedica_entity = len(entitys_dic)

# Try align entity with Wikidata
for entity in entitys_dic:
    print()
    print("="*10 + " " + entity['name'] + " " + "="*10)
    print(entity['uri'])


    # use Wikidata_id property
    request_prop_Wikidata_id = """
        PREFIX property: <{}>
        SELECT ?wikidata_id
            WHERE {{
                ?el property:Wikidata_id ?wikidata_id .
            }}
    """.format(PREFIX_PROPERTY)

    prop_Wikidata_id = make_sparql_request_wikimedica(
        entity['uri'],
        request_prop_Wikidata_id
    )

    if len(prop_Wikidata_id) == 1:
        nb_align_by_property = nb_align_by_property + 1
        entity.update(prop_Wikidata_id[0])
        print("align with 'property:Wikidata_id'")
    elif len(prop_Wikidata_id) > 1:
        print("Attention : il y a plus de 1 lien 'property:wikidata_id'")


    # use name of entity
    else:
        request_search_by_name = """
        SELECT ?entity
        WHERE {{
            ?entity rdfs:label ?entityLabel;
            wdt:P31 ?type.
            VALUES ?entityLabel {{"{}"@fr "{}"@fr}}
            VALUES ?type {{wd:Q12136 wd:Q1441305 wd:Q169872}}
        }}
        """.format((entity['name']).lower(), (entity['name']).capitalize())

        list_wikidata_pages_uri = (make_sparql_request_endpoint(
            WIKIDATA_SPARQL_ENDPOINT,
            request_search_by_name
        ))['results']['bindings']

        if len(list_wikidata_pages_uri) == 1:
            nb_align_by_name = nb_align_by_name + 1
            wikidata_uri = list_wikidata_pages_uri[0]
            entity['wikidata_id'] = rdflib.term.URIRef(wikidata_uri)
            print("align with 'name', 1 response")
            print(wikidata_uri['entity']['value'])
        elif len(list_wikidata_pages_uri) > 1:
            nb_align_conflict = nb_align_conflict + 1
            print("align with 'name', multi response")
            print(list_wikidata_pages_uri)

print()
print("nb_wikimedica_entity : ", nb_wikimedica_entity)
print("nb_align_by_property : ", nb_align_by_property)
print("nb_align_by_name : ", nb_align_by_name)
print("nb_align_conflict : ", nb_align_conflict)