def test_graph():
_check_jena()
t = SyntaxTree(TEST_SAF, sentence_id=1)
g = t.get_graphviz()
assert_equal(
set(g.edges()), {(u"t_3_Mary", u"t_2_marry"), (u"t_1_John", u"t_2_marry"), (u"t_4_little", u"t_1_John")}
)
assert_in("lemma: John", g.get_node("t_1_John").attr["label"])
# Can we 'gray out' syntax relations after applying rules
t.apply_ruleset(TEST_RULES)
t.apply_lexicon([{"lexclass": "man", "lemma": "john"}])
g = t.get_graphviz() # triple_args_function=VIS_GREY_REL)
assert_equal(
set(g.edges()),
{
(u"t_3_Mary", u"t_2_marry"),
(u"t_1_John", u"t_2_marry"),
(u"t_4_little", u"t_1_John"),
(u"t_1_John", u"t_3_Mary"),
},
)
# assert_false(g.get_edge(u't_1_John', u't_3_Mary').attr.get("color"))
# assert_equal(g.get_edge(u't_1_John', u't_2_marry').attr.get("color"),
# "grey")
# can we draw it? can't check output, but we can check for errors
with tempfile.NamedTemporaryFile() as f:
g.draw(f.name, prog="dot")
g.draw("/tmp/test.png", prog="dot")
def test_load():
_check_jena()
t = SyntaxTree(TEST_SAF, sentence_id=2)
triples = list(t.get_triples())
assert_equal(len(triples), 1)
assert_equal(triples[0].predicate, "rel_nsubj")
assert_equal(triples[0].subject.lemma, "it")
assert_equal(triples[0].object.lemma, "rain")
def test_rules():
_check_jena()
t = SyntaxTree(TEST_SAF, sentence_id=1)
t.apply_ruleset(TEST_RULES)
triples = [tr for tr in t.get_triples() if tr.predicate == "marry"]
assert_equal(len(triples), 1)
assert_equal(triples[0].subject.lemma, "John")
assert_equal(triples[0].object.lemma, "Mary")
def test_lexicon():
_check_jena()
t = SyntaxTree(TEST_SAF, sentence_id=1)
t.apply_lexicon(TEST_RULES["lexicon"])
classes = {k.replace(NS_AMCAT, ":"): v.get("lexclass") for (k, v) in t.get_tokens().iteritems()}
assert_equal(
classes, {":t_2_marry": ["marry"], ":t_3_Mary": ["person"], ":t_1_John": ["person"], ":t_4_little": None}
)
def test_get_triples():
_check_jena()
t = SyntaxTree(TEST_SAF, sentence_id=1)
t.apply_ruleset(TEST_RULES)
rels = list(t.get_relations())
marry = dict(predicate="marry", object=3, object_nodes=[3], subject=1, subject_nodes=[1, 4])
assert_equal(rels, [marry])
# add new relation, see if percolation is 'blocked' by relation
t.apply_rule({"condition": "?x :rel_dobj ?y", "insert": "?y :test ?x"})
rels = sorted(t.get_relations(), key=lambda rel: rel["predicate"])
assert_equal(rels[0], marry)
assert_equal(rels[1], dict(predicate="test", object=3, object_nodes=[3], subject=2, subject_nodes=[2]))
def test_rules_multiple():
"""Can we apply rules to multiple sentences at once"""
_check_jena()
triples_single = set()
for sid in {t["sentence"] for t in TEST_SAF["tokens"]}:
t = SyntaxTree(TEST_SAF, sentence_id=sid)
t.apply_ruleset(TEST_RULES)
triples_single |= set((s.id, p, o.id) for (s, p, o) in t.get_triples())
t = SyntaxTree(TEST_SAF)
t.apply_ruleset(TEST_RULES)
triples_all = set((s.id, p, o.id) for (s, p, o) in t.get_triples())
assert_equal(triples_single, triples_all)
def test_sparql():
from tools import _check_jena
_check_jena()
r = SparqlRunner()
data = """
@prefix vCard: <http://www.w3.org/2001/vcard-rdf/3.0#> .
@prefix rdf: <http://www.w3.org/1999/02/22-rdf-syntax-ns#> .
<http://somewhere/MattJones/> vCard:FN "Matt Jones" .
<http://somewhere/MattJones/> vCard:N _:b0 .
_:b0 vCard:Family "Jones" .
_:b0 vCard:Given "Matthew" .
"""
update = '''PREFIX vCard: <http://www.w3.org/2001/vcard-rdf/3.0#>
DELETE {?x vCard:FN "Matt Jones"}
INSERT {?x vCard:FN "Matty" }
WHERE {?x vCard:FN "Matt Jones"}'''
g = r.run(data, [update])
rels = {tuple(map(str, spo)) for spo in list(g)}
MATT, FN = 'http://somewhere/MattJones/', 'http://www.w3.org/2001/vcard-rdf/3.0#FN'
# insert worked?
assert_in((MATT, FN, 'Matty'), rels)
# delete worked?
assert_not_in((MATT, FN, 'Matt Jones'), rels)
# test rdflib Graph input
Graph().parse(data=data, format='n3')
g = r.run(data, [update])
assert_in((MATT, FN, 'Matty'), rels)
# test error handling, is the sparql exception reproduced
try:
r.run(data, ["THIS IS NOT VALID SPARQL"])
except Exception, e:
assert_in("Lexical error", str(e))