def test_Node_isgpred_realpred_node(self):
gnode = Node(pred='the_q')
realnode = Node(pred='_cat_n')
self.assertTrue(gnode.is_gpred_node)
self.assertTrue(realnode.is_realpred_node)
self.assertFalse(gnode.is_realpred_node)
self.assertFalse(realnode.is_gpred_node)
def test_Node_str(self):
node = Node()
self.assertEqual(str(node), "None")
node = Node(nodeid=2,
pred='_dog_n_1',
sortinfo=dict(cvarsort='i', pers='3', num='sg', ind='+'),
carg='Pat')
self.assertEqual(str(node), '_dog_n_1(Pat) x[pers=3, num=sg, ind=+]')
def the_dog_chases_the_cat_and_the_mouse():
return DictDmrs(nodes=[
Node(nodeid=1, pred=RealPred('the', 'q')),
Node(nodeid=2,
pred=RealPred('dog', 'n', '1'),
sortinfo=InstanceSortinfo(pers='3', num='sg', ind='+')),
Node(nodeid=3,
pred=RealPred('chase', 'v', '1'),
sortinfo=EventSortinfo(sf='prop', tense='pres',
mood='indicative')),
Node(nodeid=4, pred=RealPred('the', 'q')),
Node(nodeid=5,
pred=RealPred('cat', 'n', '1'),
sortinfo=InstanceSortinfo(pers='3', num='sg', ind='+')),
Node(nodeid=6, pred=GPred('udef_q')),
Node(nodeid=7,
pred=RealPred('and', 'c'),
sortinfo=InstanceSortinfo(pers='3', num='pl')),
Node(nodeid=8, pred=RealPred('the', 'q')),
Node(nodeid=9,
pred=RealPred('mouse', 'n', '1'),
sortinfo=InstanceSortinfo(pers='3', num='sg', ind='+'))
],
links=[
Link(start=1, end=2, rargname='RSTR', post='H'),
Link(start=3, end=2, rargname='ARG1', post='NEQ'),
Link(start=3, end=7, rargname='ARG2', post='NEQ'),
Link(start=4, end=5, rargname='RSTR', post='H'),
Link(start=6, end=7, rargname='RSTR', post='H'),
Link(start=7, end=5, rargname='L-INDEX', post='NEQ'),
Link(start=7, end=9, rargname='R-INDEX', post='NEQ'),
Link(start=8, end=9, rargname='RSTR', post='H')
],
index=3,
top=3)
def the_mouse():
dmrs = DictDmrs(surface='the mouse')
dmrs.add_node(Node(nodeid=1, pred=RealPred('the', 'q'), cfrom=0, cto=3))
dmrs.add_node(
Node(nodeid=2,
pred=RealPred('mouse', 'n', '1'),
cfrom=4,
cto=9,
sortinfo=InstanceSortinfo(pers='3', num='sg', ind='+')))
dmrs.add_link(Link(start=1, end=2, rargname='RSTR', post='H'))
return dmrs
def dog_cat():
dmrs = DictDmrs(surface='dog cat')
dmrs.add_node(
Node(pred=RealPred('dog', 'n', '1'),
cfrom=0,
cto=3,
sortinfo=InstanceSortinfo(pers='3', num='sg', ind='+')))
dmrs.add_node(
Node(pred=RealPred('cat', 'n', '1'),
cfrom=4,
cto=7,
sortinfo=InstanceSortinfo(pers='3', num='sg', ind='+')))
return dmrs
def test_Node_init(self):
node = Node(
nodeid=13,
pred='the_q',
surface='cat',
base='x',
cfrom=23,
cto=27,
carg='Kim',
)
self.assertEqual(node.nodeid, 13)
self.assertEqual(node.surface, 'cat')
self.assertEqual(node.base, 'x')
self.assertEqual(node.cfrom, 23)
self.assertEqual(node.cto, 27)
# Incorrect span
with self.assertRaises(PydmrsValueError):
Node(cfrom=22, cto=7)
self.assertEqual(node.carg, 'Kim')
# Fix carg with "".
self.assertEqual(Node(carg='"Kim"').carg, 'Kim')
# Unaccounted " in carg
with self.assertRaises(PydmrsValueError):
Node(carg='"Kim')
# String pred.
self.assertEqual(node.pred, GPred('the_q'))
# Other pred
self.assertEqual(Node(pred=GPred('the_q')).pred, GPred('the_q'))
# Allow None for sortinfo.
self.assertEqual(Node().sortinfo, None)
# Dict sortinfo
self.assertEqual(
Node(sortinfo={
'cvarsort': 'i',
'pers': '3'
}).sortinfo, InstanceSortinfo(pers='3'))
# Sortinfo sortinfo
self.assertEqual(
Node(sortinfo=InstanceSortinfo(pers='3')).sortinfo,
InstanceSortinfo(pers='3'))
# List sortinfo
self.assertEqual(
Node(sortinfo=[('cvarsort', 'i'), ('pers', '3')]).sortinfo,
InstanceSortinfo(pers='3'))
# But nothing else.
with self.assertRaises(PydmrsTypeError):
Node(sortinfo="x[pers=3, num=sg, ind=+]")
def test_get_matched_subgraph(self):
match = general_matching.find_best_matches(self.cat_dmrs,
self.small_dmrs)[0]
subgraph = general_matching.get_matched_subgraph(
self.small_dmrs, match)
expected = DictDmrs(
nodes=[
Node(nodeid=4, pred=RealPred('the', 'q')),
Node(nodeid=5,
pred=RealPred('cat', 'n', '1'),
sortinfo=InstanceSortinfo(pers='3', num='sg', ind='+'))
],
links=[Link(start=4, end=5, rargname='RSTR', post='H')])
self.assertListEqual(subgraph.nodes, expected.nodes)
self.assertListEqual(subgraph.links, expected.links)
def test_get_matching_nodeids(self):
# Match "the cat" onto "the dog chases the cat" (exact fit, only one match)
matches1 = aligned_matching.get_matching_nodeids(
self.the_cat, self.the_dog_chases_the_cat)
self.assertEqual(len(matches1), 1)
self.assertCountEqual(matches1[0], [(2, 5), (1, 4)])
# all_surface = True
all_matches1 = aligned_matching.get_matching_nodeids(
self.the_cat, self.the_dog_chases_the_cat, all_surface=True)
# The same as earlier
self.assertListEqual(matches1[0], all_matches1[0])
# Extra surface nodes: between dog and cat
all_matches1 = aligned_matching.get_matching_nodeids(
self.dog_cat, self.the_dog_chases_the_cat, all_surface=True)
self.assertCountEqual(all_matches1[0], [(2, 5), (1, 2), (None, 3),
(None, 4)])
# Match "the dog chases the cat" onto "the cat chases the dog" (inexact fit)
matches2 = aligned_matching.get_matching_nodeids(
self.the_dog_chases_the_cat, self.the_cat_chases_the_dog)
# Two options: "the dog" matches or "the cat" matches, 'chases' doesn't because it's not part of the longest match
self.assertEqual(len(matches2), 2)
self.assertCountEqual(matches2, [[(5, 2), (4, 1)], [(2, 5), (1, 4)]])
# No match found
matches = aligned_matching.get_matching_nodeids(
self.the_mouse, self.dog_cat)
self.assertListEqual(matches, [])
# Should be the same as 'the cat'.
mixed_cat = ListDmrs(surface='the cat')
mixed_cat.add_node(
Node(nodeid=2,
pred=RealPred('cat', 'n', '1'),
cfrom=4,
cto=7,
sortinfo=InstanceSortinfo(pers='3', num='sg', ind='+')))
mixed_cat.add_node(
Node(nodeid=1, pred=RealPred('the', 'q'), cfrom=0, cto=3))
mixed_cat.add_link(Link(start=1, end=2, rargname='RSTR', post='H'))
mixed = aligned_matching.get_matching_nodeids(
mixed_cat, self.the_dog_chases_the_cat)
self.assertListEqual(mixed, matches1)
def the_cat_chases_the_dog():
return DictDmrs(surface='the cat chases the dog',
nodes=[
Node(nodeid=1,
pred=RealPred('the', 'q'),
cfrom=0,
cto=3),
Node(nodeid=2,
pred=RealPred('cat', 'n', '1'),
cfrom=4,
cto=7,
sortinfo=InstanceSortinfo(pers='3',
num='sg',
ind='+')),
Node(nodeid=3,
pred=RealPred('chase', 'v', '1'),
cfrom=8,
cto=14,
sortinfo=EventSortinfo(sf='prop',
tense='pres',
mood='indicative')),
Node(nodeid=4,
pred=RealPred('the', 'q'),
cfrom=15,
cto=18),
Node(nodeid=5,
pred=RealPred('dog', 'n', '1'),
cfrom=19,
cto=22,
sortinfo=InstanceSortinfo(pers='3',
num='sg',
ind='+'))
],
links=[
Link(start=1, end=2, rargname='RSTR', post='H'),
Link(start=3, end=2, rargname='ARG1', post='NEQ'),
Link(start=3, end=5, rargname='ARG2', post='NEQ'),
Link(start=4, end=5, rargname='RSTR', post='H')
],
index=3,
top=3)
def loads_xml(bytestring,
encoding=None,
cls=ListDmrs,
convert_legacy_prontype=True,
**kwargs):
"""
Currently processes "<dmrs>...</dmrs>"
To be updated for "<dmrslist>...</dmrslist>"...
Expects a bytestring; to load from a string instead, specify encoding
Produces a ListDmrs by default; for a different type, specify cls
"""
if encoding:
bytestring = bytestring.encode(encoding)
xml = ET.XML(bytestring)
dmrs = cls(**kwargs)
dmrs.cfrom = int(xml.get('cfrom')) if 'cfrom' in xml.attrib else None
dmrs.cto = int(xml.get('cto')) if 'cto' in xml.attrib else None
dmrs.surface = xml.get('surface')
dmrs.ident = int(xml.get('ident')) if 'ident' in xml.attrib else None
index_id = int(xml.get('index')) if 'index' in xml.attrib else None
top_id = None
for elem in xml:
if elem.tag == 'node':
node = Node.from_xml(elem, convert_legacy_prontype)
dmrs.add_node(node)
elif elem.tag == 'link':
link = Link.from_xml(elem)
if link.start == 0:
top_id = link.end
else:
dmrs.add_link(link)
else:
raise PydmrsValueError(elem.tag)
if top_id:
dmrs.top = dmrs[top_id]
if index_id:
dmrs.index = dmrs[index_id]
return dmrs
def loads_xml(bytestring, encoding=None, cls=ListDmrs, convert_legacy_prontype=True, **kwargs):
"""
Currently processes "<dmrs>...</dmrs>"
To be updated for "<dmrslist>...</dmrslist>"...
Expects a bytestring; to load from a string instead, specify encoding
Produces a ListDmrs by default; for a different type, specify cls
"""
if encoding:
bytestring = bytestring.encode(encoding)
xml = ET.XML(bytestring)
dmrs = cls(**kwargs)
dmrs.cfrom = int(xml.get('cfrom')) if 'cfrom' in xml.attrib else None
dmrs.cto = int(xml.get('cto')) if 'cto' in xml.attrib else None
dmrs.surface = xml.get('surface')
dmrs.ident = int(xml.get('ident')) if 'ident' in xml.attrib else None
# top may be set as a graph attribute or as a link (see below)
top_id = int(xml.get('top')) if 'top' in xml.attrib else None
index_id = int(xml.get('index')) if 'index' in xml.attrib else None
for elem in xml:
if elem.tag == 'node':
node = Node.from_xml(elem, convert_legacy_prontype)
dmrs.add_node(node)
elif elem.tag == 'link':
link = Link.from_xml(elem)
if link.start == 0:
# this would overwrite any graph-level top attribute
# (see above), but let's assume we won't encounter
# both in the same graph
top_id = link.end
else:
dmrs.add_link(link)
else:
raise PydmrsValueError(elem.tag)
if top_id:
dmrs.top = dmrs[top_id]
if index_id:
dmrs.index = dmrs[index_id]
return dmrs
def test_Node_eq(self):
# Unspecified nodes are always equal.
node1 = Node()
node2 = Node()
self.assertEqual(node1, node2)
sortinfo1 = {'cvarsort': 'e', 'tense': 'past'}
sortinfo2 = {'cvarsort': 'e', 'tense': 'pres'}
# Two nodes are equal if they have the same pred, sortinfo and carg,
# even if all the other elements are different
node1 = Node(nodeid=23,
pred='the_q',
sortinfo=sortinfo1,
cfrom=2,
cto=22,
carg='Kim',
surface='cat',
base='x')
node2 = Node(nodeid=25,
pred='the_q',
sortinfo=sortinfo1,
cfrom=15,
carg='Kim',
surface='mad',
base='w')
self.assertEqual(node1, node2)
# Different carg
node2 = Node(pred='the_q', sortinfo=sortinfo1, carg='Jane')
self.assertNotEqual(node1, node2)
# Different pred
node2 = Node(pred='_smile_v', sortinfo=sortinfo1, carg='Kim')
self.assertNotEqual(node1, node2)
# Different sortinfo.
node2 = Node(pred='_the_q', sortinfo=sortinfo2, carg='Kim')
self.assertNotEqual(node1, node2)
def test_Node_span(self):
node = Node(cfrom=2, cto=15)
self.assertEqual(node.span, (2, 15))
def test_Node_underspecification(self):
with self.assertRaises(TypeError):
Node(pred='_the_q').is_more_specific(4)
# complete underspecification
self.assertFalse(Node().is_more_specific(Node()))
self.assertFalse(Node().is_less_specific(Node()))
# pred underspecification
self.assertFalse(Node(pred=Pred()).is_more_specific(Node()))
self.assertTrue(Node(pred=Pred()).is_less_specific(Node()))
self.assertTrue(Node().is_more_specific(Node(pred=Pred())))
self.assertFalse(Node().is_less_specific(Node(pred=Pred())))
self.assertFalse(Node(pred=Pred()).is_more_specific(Node(pred=Pred())))
self.assertFalse(Node(pred=Pred()).is_less_specific(Node(pred=Pred())))
self.assertFalse(
Node(pred=Pred()).is_more_specific(Node(pred=GPred(name='abc'))))
self.assertTrue(
Node(pred=Pred()).is_less_specific(Node(pred=GPred(name='abc'))))
self.assertTrue(
Node(pred=GPred(name='abc')).is_more_specific(Node(pred=Pred())))
self.assertFalse(
Node(pred=GPred(name='abc')).is_less_specific(Node(pred=Pred())))
# carg underspecification
self.assertFalse(Node(carg='?').is_more_specific(Node()))
self.assertTrue(Node(carg='?').is_less_specific(Node()))
self.assertTrue(Node().is_more_specific(Node(carg='?')))
self.assertFalse(Node().is_less_specific(Node(carg='?')))
self.assertFalse(Node(carg='?').is_more_specific(Node(carg='?')))
self.assertFalse(Node(carg='?').is_less_specific(Node(carg='?')))
self.assertFalse(Node(carg='?').is_more_specific(Node(carg='abc')))
self.assertTrue(Node(carg='?').is_less_specific(Node(carg='abc')))
self.assertTrue(Node(carg='abc').is_more_specific(Node(carg='?')))
self.assertFalse(Node(carg='abc').is_less_specific(Node(carg='?')))
# sortinfo underspecification
self.assertFalse(Node(sortinfo=Sortinfo()).is_more_specific(Node()))
self.assertTrue(Node(sortinfo=Sortinfo()).is_less_specific(Node()))
self.assertTrue(Node().is_more_specific(Node(sortinfo=Sortinfo())))
self.assertFalse(Node().is_less_specific(Node(sortinfo=Sortinfo())))
self.assertFalse(
Node(sortinfo=Sortinfo()).is_more_specific(
Node(sortinfo=Sortinfo())))
self.assertFalse(
Node(sortinfo=Sortinfo()).is_less_specific(
Node(sortinfo=Sortinfo())))
self.assertFalse(
Node(sortinfo=Sortinfo()).is_more_specific(
Node(sortinfo=EventSortinfo(sf='abc'))))
self.assertTrue(
Node(sortinfo=Sortinfo()).is_less_specific(
Node(sortinfo=EventSortinfo(sf='abc'))))
self.assertTrue(
Node(sortinfo=EventSortinfo(sf='abc')).is_more_specific(
Node(sortinfo=Sortinfo())))
self.assertFalse(
Node(sortinfo=EventSortinfo(sf='abc')).is_less_specific(
Node(sortinfo=Sortinfo())))
# mixed specification
self.assertFalse(Node(pred=Pred()).is_more_specific(Node(carg='?')))
self.assertFalse(Node(pred=Pred()).is_less_specific(Node(carg='?')))
self.assertFalse(
Node(pred=Pred()).is_more_specific(Node(sortinfo=Sortinfo())))
self.assertFalse(
Node(pred=Pred()).is_less_specific(Node(sortinfo=Sortinfo())))
self.assertFalse(
Node(carg='?').is_more_specific(Node(sortinfo=Sortinfo())))
self.assertFalse(
Node(carg='?').is_less_specific(Node(sortinfo=Sortinfo())))
def predsort():
dmrs = DictDmrs()
dmrs.add_node(
Node(pred=Pred(),
sortinfo=Sortinfo())) # underspecified predicate and sortinfo
return dmrs
def noun():
dmrs = DictDmrs()
dmrs.add_node(
Node(pred=RealPred('?', 'n', 'unknown'),
sortinfo=Sortinfo())) # underspecified noun and sortinfo
return dmrs
def _parse_node(string, nodeid, queries, equalities, anchors, sortinfo_classes,
sortinfo_shortforms):
m = string.find('(')
if m < 0:
m = string.find(' ')
if m < 0:
l = string.find(':')
else:
l = string.find(':', 0, m)
if l < 0:
ref_ids = None
l = 0
else:
ref_ids = string[:l]
l += 1
while string[l] == ' ':
l += 1
if string[l:l + 4] == 'node' and (len(string) - l == 4
or string[l + 4] in special_values):
value = _parse_value(string[l + 4:], None, queries, equalities,
(lambda matching, dmrs: dmrs[matching[nodeid]]))
assert not value
pred = Pred()
carg = '?'
sortinfo = Sortinfo()
ref_name = 'node'
elif m < 0:
pred, ref_name = _parse_pred(string[l:], nodeid, queries, equalities)
carg = None
sortinfo = None
else:
pred, ref_name = _parse_pred(string[l:m], nodeid, queries, equalities)
if string[m] == '(':
r = string.index(')', m)
if string[m + 1] == '"' and string[r - 1] == '"':
carg = string[m + 2:r - 1]
else:
carg = string[m + 1:r]
assert '"' not in carg
carg = _parse_value(
carg, '?', queries, equalities,
(lambda matching, dmrs: dmrs[matching[nodeid]].carg))
m = r + 1
else:
carg = None
if m < len(string) and string[m] == ' ':
while string[m] == ' ':
m += 1
sortinfo = _parse_sortinfo(string[m:], nodeid, queries, equalities,
sortinfo_classes, sortinfo_shortforms)
else:
sortinfo = None
if not ref_ids:
ref_ids = None
node = Node(nodeid, pred, sortinfo=sortinfo, carg=carg)
else:
if ref_ids[0] == '[' and ref_ids[-1] == ']':
ref_ids = ref_ids[1:-1].split(',')
node = AnchorNode(anchors=ref_ids,
nodeid=nodeid,
pred=pred,
sortinfo=sortinfo,
carg=carg)
elif ref_ids[0] == '(' and ref_ids[-1] == ')':
ref_ids = ref_ids[1:-1].split(',')
node = OptionalNode(anchors=ref_ids,
nodeid=nodeid,
pred=pred,
sortinfo=sortinfo,
carg=carg)
elif ref_ids[0] == '{' and ref_ids[-1] == '}':
ref_ids = ref_ids[1:-1].split(',')
node = SubgraphNode(anchors=ref_ids,
nodeid=nodeid,
pred=pred,
sortinfo=sortinfo,
carg=carg)
else:
ref_ids = ref_ids.split(',')
node = Node(nodeid=nodeid, pred=pred, sortinfo=sortinfo, carg=carg)
for ref_id in ref_ids:
assert ref_id not in anchors, 'Reference ids have to be unique.'
anchors[ref_id] = node
return node, ref_ids, ref_name
def the():
dmrs = DictDmrs()
dmrs.add_node(Node(pred=RealPred('the', 'q'))) # node id set automatically
return dmrs