Building a Knowledge Graph

from rdflib import Graph, Namespace, URIRef, Literal, BNode
from rdflib.namespace import FOAF, RDF, RDFS, XSD, OWL
import pandas as pd

# Define namespaces
EX   = Namespace("http://example.org/")
ORG  = Namespace("http://www.w3.org/ns/org#")
SCHEMA = Namespace("https://schema.org/")

g = Graph()
g.bind("ex", EX)
g.bind("foaf", FOAF)
g.bind("org", ORG)
g.bind("schema", SCHEMA)

# ── 1. Define schema (T-Box) ──────────────────────────
g.add((EX.Employee, RDF.type, OWL.Class))
g.add((EX.Employee, RDFS.subClassOf, FOAF.Person))
g.add((EX.Employee, RDFS.label, Literal("Employee")))

g.add((EX.worksAt, RDF.type, OWL.ObjectProperty))
g.add((EX.worksAt, RDFS.domain, EX.Employee))
g.add((EX.worksAt, RDFS.range, ORG.Organization))

# ── 2. Load raw data ──────────────────────────────────
df = pd.read_csv("employees.csv")

company_cache: dict[str, URIRef] = {}
city_cache: dict[str, URIRef] = {}

for _, row in df.iterrows():
    # Person URI
    slug = row["name"].lower().replace(" ", "_")
    person = EX[f"person/{slug}"]

    g.add((person, RDF.type, EX.Employee))
    g.add((person, FOAF.name, Literal(row["name"])))
    g.add((person, FOAF.age, Literal(int(row["age"]), datatype=XSD.integer)))

    # Company URI (deduplication via cache)
    company_name = row["company"]
    if company_name not in company_cache:
        co_slug = company_name.lower().replace(" ", "_")
        company = EX[f"company/{co_slug}"]
        company_cache[company_name] = company
        g.add((company, RDF.type, ORG.Organization))
        g.add((company, FOAF.name, Literal(company_name)))
    company = company_cache[company_name]

    # City URI
    city_name = row["city"]
    if city_name not in city_cache:
        city_slug = city_name.lower().replace(" ", "_")
        city = EX[f"city/{city_slug}"]
        city_cache[city_name] = city
        g.add((city, RDF.type, SCHEMA.City))
        g.add((city, FOAF.name, Literal(city_name)))
    city = city_cache[city_name]

    # Relationships
    g.add((person, EX.worksAt, company))
    g.add((company, SCHEMA.location, city))

print(f"Graph has {len(g)} triples")

# ── 3. Serialize ──────────────────────────────────────
g.serialize(destination="knowledge_graph.ttl", format="turtle")
g.serialize(destination="knowledge_graph.jsonld", format="json-ld")

# ── 4. Query ──────────────────────────────────────────
result = g.query("""
    SELECT ?name ?company WHERE {
        ?p foaf:name ?name ;
           ex:worksAt ?c .
        ?c foaf:name ?company .
    }
""", initNs={"foaf": FOAF, "ex": EX})

for row in result:
    print(f"{row.name} works at {row.company}")

from SPARQLWrapper import SPARQLWrapper, JSON

def link_city_to_wikidata(city_name: str) -> str | None:
    """Look up a city in Wikidata and return its QID."""
    sparql = SPARQLWrapper("https://query.wikidata.org/sparql")
    sparql.setQuery(f"""
        SELECT ?item WHERE {{
            ?item wdt:P31 wd:Q515 ;      # instance of: city
                  rdfs:label "{city_name}"@en .
        }}
        LIMIT 1
    """)
    sparql.setReturnFormat(JSON)
    results = sparql.query().convert()

    if results["results"]["bindings"]:
        uri = results["results"]["bindings"][0]["item"]["value"]
        return uri  # e.g. "https://www.wikidata.org/entity/Q84"
    return None

# Add owl:sameAs link to Wikidata
from rdflib import OWL, URIRef

wikidata_uri = link_city_to_wikidata("London")
if wikidata_uri:
    g.add((EX["city/london"], OWL.sameAs, URIRef(wikidata_uri)))
    print(f"Linked London → {wikidata_uri}")