For example: <Person:Einstein> <property:birthPlace> <Place:Ulm>. Nodes are either URIs (resources) or literals (strings, numbers). Collections of triples form a directed labeled graph.
OWL (Web Ontology Language) adds axioms: owl:subClassOf, owl:equivalentClass, owl:inverseOf. A reasoner can infer new triples: if A subClassOf B and x type A, then x type B.
from rdflib import Graph
import pandas as pd
# Reload the graph
g = Graph()
g.parse("research_graph.ttl", format="turtle")
# -----------------------------------------------------------------# Query 1: All papers with author names and years# -----------------------------------------------------------------
q1 = """
PREFIX dcterms: <http://purl.org/dc/terms/>
PREFIX foaf: <http://xmlns.com/foaf/0.1/>
PREFIX bibo: <http://purl.org/ontology/bibo/>
SELECT ?title ?author_name ?year
WHERE {
?paper a bibo:AcademicArticle ;
dcterms:title ?title ;
dcterms:date ?year ;
dcterms:creator ?author .
?author foaf:name ?author_name .
}
ORDER BY ?year ?title
"""
results1 = g.query(q1)
df1 = pd.DataFrame(results1, columns=["title", "author_name", "year"])
print("=== Query 1: Papers and Authors ===")
print(df1.to_string(index=False))
# -----------------------------------------------------------------# Query 2: Co-author pairs# -----------------------------------------------------------------
q2 = """
PREFIX dcterms: <http://purl.org/dc/terms/>
PREFIX foaf: <http://xmlns.com/foaf/0.1/>
PREFIX bibo: <http://purl.org/ontology/bibo/>
SELECT DISTINCT ?author1 ?author2 (COUNT(?paper) AS ?collab_count)
WHERE {
?paper a bibo:AcademicArticle ;
dcterms:creator ?a1 ;
dcterms:creator ?a2 .
?a1 foaf:name ?author1 .
?a2 foaf:name ?author2 .
FILTER(?a1 < ?a2)
}
GROUP BY ?author1 ?author2
ORDER BY DESC(?collab_count)
"""
results2 = g.query(q2)
df2 = pd.DataFrame(results2, columns=["author1", "author2", "collaborations"])
print("\n=== Query 2: Co-author Pairs ===")
print(df2.to_string(index=False))
# -----------------------------------------------------------------# Query 3: Citation network (who cites whom via paper author)# -----------------------------------------------------------------
q3 = """
PREFIX dcterms: <http://purl.org/dc/terms/>
PREFIX bibo: <http://purl.org/ontology/bibo/>
PREFIX foaf: <http://xmlns.com/foaf/0.1/>
SELECT ?citing_paper_title ?cited_paper_title
WHERE {
?citing_paper bibo:cites ?cited_paper .
?citing_paper dcterms:title ?citing_paper_title .
?cited_paper dcterms:title ?cited_paper_title .
}
"""
results3 = g.query(q3)
df3 = pd.DataFrame(results3, columns=["citing", "cited"])
print("\n=== Query 3: Citation Relationships ===")
print(df3.to_string(index=False))
# -----------------------------------------------------------------# Query 4: Concept co-occurrence in papers# -----------------------------------------------------------------
q4 = """
PREFIX schema: <https://schema.org/>
PREFIX dcterms: <http://purl.org/dc/terms/>
PREFIX bibo: <http://purl.org/ontology/bibo/>
SELECT ?c1_name ?c2_name (COUNT(?paper) AS ?co_count)
WHERE {
?paper a bibo:AcademicArticle ;
schema:about ?c1 ;
schema:about ?c2 .
?c1 schema:name ?c1_name .
?c2 schema:name ?c2_name .
FILTER(?c1 < ?c2)
}
GROUP BY ?c1_name ?c2_name
HAVING (?co_count >= 1)
ORDER BY DESC(?co_count)
"""
results4 = g.query(q4)
df4 = pd.DataFrame(results4, columns=["concept1", "concept2", "co_count"])
print("\n=== Query 4: Concept Co-occurrence ===")
print(df4.to_string(index=False))
Step 3: Wikidata SPARQL Queries
import requests
import pandas as pd
import time
import os
WIKIDATA_ENDPOINT = "https://query.wikidata.org/sparql"defwikidata_sparql(query, timeout=30):
"""Execute a SPARQL query against Wikidata.
Args:
query: SPARQL query string
timeout: request timeout in seconds
Returns:
DataFrame with results, or synthetic fallback on error
"""
headers = {
"User-Agent": "ResearchBot/1.0 (academic research)",
"Accept": "application/sparql-results+json",
}
try:
response = requests.get(
WIKIDATA_ENDPOINT,
params={"query": query, "format": "json"},
headers=headers,
timeout=timeout
)
response.raise_for_status()
data = response.json()
vars_ = data["head"]["vars"]
rows = []
for binding in data["results"]["bindings"]:
row = {v: binding.get(v, {}).get("value", "") for v in vars_}
rows.append(row)
return pd.DataFrame(rows), Trueexcept Exception as e:
print(f"Wikidata query failed: {e}")
return pd.DataFrame(), False# Query: Top universities founded before 1600 with their location
query_universities = """
SELECT ?university ?universityLabel ?founded ?countryLabel
WHERE {
?university wdt:P31 wd:Q3918 . # instance of: university
?university wdt:P571 ?founded . # founded date
?university wdt:P17 ?country . # country
FILTER(YEAR(?founded) < 1600)
SERVICE wikibase:label { bd:serviceParam wikibase:language "en" }
}
ORDER BY ?founded
LIMIT 20
"""
df_univ, success = wikidata_sparql(query_universities)
if success andlen(df_univ) > 0:
print("=== Historical Universities (from Wikidata) ===")
print(df_univ[["universityLabel", "founded", "countryLabel"]].to_string(index=False))
else:
# Synthetic fallbackprint("=== Historical Universities (synthetic data) ===")
synthetic = pd.DataFrame({
"universityLabel": ["University of Bologna", "University of Oxford",
"University of Cambridge", "University of Salamanca",
"University of Paris"],
"founded": ["1088", "1096", "1209", "1218", "1257"],
"countryLabel": ["Italy", "UK", "UK", "Spain", "France"],
})
print(synthetic.to_string(index=False))
# Query: Nobel laureates in physics with their alma mater
query_nobel = """
SELECT ?person ?personLabel ?year ?universityLabel
WHERE {
?person wdt:P166 wd:Q38104 . # award: Nobel Prize in Physics
?person p:P166 ?statement .
?statement pq:P585 ?year .
OPTIONAL { ?person wdt:P69 ?university . }
SERVICE wikibase:label { bd:serviceParam wikibase:language "en" }
}
ORDER BY DESC(?year)
LIMIT 20
"""
df_nobel, success = wikidata_sparql(query_nobel)
if success andlen(df_nobel) > 0:
print("\n=== Recent Nobel Physics Laureates ===")
print(df_nobel[["personLabel", "year"]].head(10).to_string(index=False))
else:
print("\n(Wikidata offline – skipping Nobel query)")
# -----------------------------------------------------------------# Entity lookup by ORCID# -----------------------------------------------------------------deflookup_researcher_wikidata(orcid_id):
"""Find Wikidata entity for a researcher by ORCID.
Args:
orcid_id: ORCID string (e.g. "0000-0002-1825-0097")
Returns:
dict with Wikidata QID and labels, or empty dict on failure
"""
query = f"""
SELECT ?person ?personLabel ?employerLabel
WHERE {{
?person wdt:P496 "{orcid_id}" .
OPTIONAL {{ ?person wdt:P108 ?employer . }}
SERVICE wikibase:label {{ bd:serviceParam wikibase:language "en" }}
}}
"""
df, ok = wikidata_sparql(query)
if ok andlen(df) > 0:
return df.iloc[0].to_dict()
return {}
# Example ORCID (Tim Berners-Lee: 0000-0003-1279-3709)
researcher = lookup_researcher_wikidata("0000-0003-1279-3709")
if researcher:
print(f"\nWikidata entity found: {researcher}")
else:
print("\nEntity lookup: API offline or ORCID not found in Wikidata")
Advanced Usage
Federated SPARQL Query
-- Example: Link local data with Wikidata via federated query
-- (Run this in a SPARQL endpoint that supports SERVICE)
PREFIX owl: <http://www.w3.org/2002/07/owl#>
PREFIX schema: <https://schema.org/>
SELECT ?localAuthor ?wd_label ?wd_birthdate
WHERE {
# Local graph
?localAuthor a schema:Person ;
owl:sameAs ?wikidataURI .
# Federated to Wikidata
SERVICE <https://query.wikidata.org/sparql> {
?wikidataURI wdt:P569 ?wd_birthdate .
SERVICE wikibase:label {
bd:serviceParam wikibase:language "en" .
?wikidataURI rdfs:label ?wd_label .
}
}
}
OWL Reasoning with rdflib + OWL-RL
from rdflib import Graph, Namespace, URIRef
from rdflib.namespace import RDF, RDFS, OWL
defapply_rdfs_closure(g):
"""Apply basic RDFS inference rules (subClassOf, subPropertyOf).
Args:
g: rdflib Graph
Returns:
g: graph with inferred triples added
"""
changed = Truewhile changed:
changed = False
new_triples = set()
# Rule: subClassOf transitivityfor s, _, sc in g.triples((None, RDFS.subClassOf, None)):
for _, _, sc2 in g.triples((sc, RDFS.subClassOf, None)):
t = (s, RDFS.subClassOf, sc2)
if t notin g:
new_triples.add(t)
# Rule: type propagation via subClassOffor s, _, type_ in g.triples((None, RDF.type, None)):
for _, _, superclass in g.triples((type_, RDFS.subClassOf, None)):
t = (s, RDF.type, superclass)
if t notin g:
new_triples.add(t)
for t in new_triples:
g.add(t)
changed = Truereturn g
# Example: Add subClassOf axioms and infer types
g_owl = Graph()
EX2 = Namespace("http://example.org/")
g_owl.bind("ex", EX2)
# Ontology axioms
g_owl.add((EX2.Professor, RDFS.subClassOf, EX2.AcademicStaff))
g_owl.add((EX2.AcademicStaff, RDFS.subClassOf, EX2.Person))
g_owl.add((EX2.alice, RDF.type, EX2.Professor))
print(f"Before inference: {len(g_owl)} triples")
g_owl = apply_rdfs_closure(g_owl)
print(f"After RDFS closure: {len(g_owl)} triples")
for s, p, o in g_owl.triples((EX2.alice, RDF.type, None)):
print(f" alice rdf:type {o.split('/')[-1]}")
Graph Visualization with NetworkX
import networkx as nx
import matplotlib.pyplot as plt
from rdflib import Graph
defrdf_to_networkx(g, predicate_filter=None):
"""Convert rdflib Graph to NetworkX DiGraph for visualization.
Args:
g: rdflib.Graph
predicate_filter: optional set of predicate URIs to include
Returns:
G: networkx DiGraph
"""
G = nx.DiGraph()
for s, p, o in g:
s_label = str(s).split("/")[-1].split("#")[-1]
p_label = str(p).split("/")[-1].split("#")[-1]
o_label = str(o).split("/")[-1].split("#")[-1]
if predicate_filter andstr(p) notin predicate_filter:
continue
G.add_edge(s_label, o_label, predicate=p_label)
return G
g2 = Graph()
g2.parse("research_graph.ttl", format="turtle")
from rdflib.namespace import DCTERMS
G_nx = rdf_to_networkx(g2, predicate_filter={
str(DCTERMS.creator),
"http://purl.org/ontology/bibo/cites",
})
print(f"Visualization graph: {G_nx.number_of_nodes()} nodes, "f"{G_nx.number_of_edges()} edges")
fig, ax = plt.subplots(figsize=(10, 7))
pos = nx.spring_layout(G_nx, seed=42)
edge_labels = {(u, v): d["predicate"] for u, v, d in G_nx.edges(data=True)}
nx.draw_networkx(G_nx, pos, ax=ax, node_color="lightblue",
node_size=1200, font_size=7, arrows=True)
nx.draw_networkx_edge_labels(G_nx, pos, edge_labels, font_size=6, ax=ax)
ax.set_title("Research Knowledge Graph")
ax.axis("off")
plt.tight_layout()
plt.savefig("knowledge_graph_viz.png", dpi=150, bbox_inches="tight")
plt.close()
print("Figure saved: knowledge_graph_viz.png")
Troubleshooting
Problem
Cause
Fix
SPARQL parse error
Missing prefix declaration
Add PREFIX declarations at top of query
rdflib parse failure
Wrong serialization format
Check file extension; specify format="turtle" or "xml"
Wikidata timeout
Complex query or busy endpoint
Add LIMIT; break into smaller queries; use OPTIONAL not required triples
Empty SPARQL results
Namespace mismatch
Verify namespace URIs match what was used when inserting data
rdflib reasoning too slow
Large graph
Use owlrl package for OWL 2 RL reasoning instead
Unicode in literals
Non-ASCII characters
rdflib handles UTF-8 natively; ensure Literal() wraps the string
