def test_amr_print_with_reentrancy():
r: Node = Node('receive-01')
w: Node = Node('we')
t: Node = Node('thing')
r2: Node = Node('remind-01')
p: Node = Node('pay-01')
h: Node = Node('hospital')
n: Node = Node('now')
a: Node = Node('already')
r.add_child(w, 'ARG0')
r.add_child(t, 'ARG1')
r.add_child(h, 'ARG2')
r.add_child(n, 'time')
r.add_child(a, 'time')
t.add_child(r2, 'ARG0-of')
r2.add_child(p, 'ARG1')
r2.add_child(w, 'ARG2')
p.add_child(w, 'ARG0')
generated_amr_str = r.amr_print_with_reentrancy()
expected_amr_str = """(r / receive-01~e.4
:ARG0 (w / we~e.0)
:ARG1 (t / thing~e.7
:ARG0-of~e.7 (r2 / remind-01~e.7
:ARG1 (p / pay-01~e.6
:ARG0 w)
:ARG2 w))
:ARG2~e.8 (h / hospital~e.10)
:time (n / now~e.2)
:time (a / already~e.3))"""
smatch = calculate_smatch(generated_amr_str, expected_amr_str)
assert smatch == 1
def test_pre_and_post_processing_eg_2():
sentence = 'It is Santorum that is the by far major nonRomney candidate and Newt would appear to be the spoiler .'
amr_str = """(a / and~e.11
:op1 (c / candidate~e.10
:ARG1-of (m / major-02~e.8
:degree (b / by-far~e.6,7))
:mod (p3 / person~e.9 :polarity -~e.9 :wiki "Mitt_Romney"~e.9
:name (n2 / name~e.9 :op1 "Romney"~e.9))
:domain~e.1,4 (p2 / person :wiki "Rick_Santorum"
:name (n / name :op1 "Santorum"~e.2)))
:op2 (a2 / appear-02~e.14
:ARG1 (s / spoil-01~e.18
:ARG0 (p4 / person :wiki "Newt_Gingrich"
:name (n3 / name :op1 "Newt"~e.12)))))"""
amr: AMR = AMR.parse_string(amr_str)
amr, new_sentence, metadata = train_pre_processing(amr, sentence)
identified_concepts = IdentifiedConcepts()
identified_concepts.create_from_amr('', amr)
add_false_root(identified_concepts)
vector_of_parents = generate_parent_list_vector(amr, identified_concepts)
post_processing_on_parent_vector(identified_concepts, vector_of_parents,
new_sentence, metadata)
relations_dict = {
('and', 'candidate'): 'op1',
('and', 'appear-02'): 'op2',
('candidate', 'major-02'): 'ARG1-of',
('candidate', 'person'): 'mod',
('major-02', 'by-far'): 'degree',
('person', '-'): 'polarity',
('person', 'Mitt_Romney'): 'wiki',
('person', 'name'): 'name',
('person', 'Santorum'): 'wiki',
('name', 'Romney'): 'op1',
('name', 'Santorum'): 'op1',
('appear-02', 'spoil-01'): 'ARG1',
('spoil-01', 'person'): 'ARG0',
('person', 'Newt'): 'wiki',
('name', 'Newt'): 'op1'
}
amr_node: Node = generate_amr_node_for_vector_of_parents(
identified_concepts, vector_of_parents, relations_dict)
generated_amr_str = amr_node.amr_print_with_reentrancy()
expected_amr_str = """(a / and~e.11
:op1 (c / candidate~e.10
:ARG1-of (m / major-02~e.8
:degree (b / by-far~e.6,7))
:mod (p3 / person~e.9 :polarity -~e.9 :wiki "Mitt_Romney"~e.9
:name (n2 / name~e.9 :op1 "Romney"~e.9))
:mod~e.1,4 (p2 / person :wiki "Santorum"
:name (n / name :op1 "Santorum"~e.2)))
:op2 (a2 / appear-02~e.14
:ARG1 (s / spoil-01~e.18
:ARG0 (p4 / person :wiki "Newt"
:name (n3 / name :op1 "Newt"~e.12)))))"""
smatch = calculate_smatch(generated_amr_str, expected_amr_str)
assert smatch == 1
def test_generate_amr_node_for_vector_of_parents_example_1():
amr_str = """(s / suppose-01~e.1
:ARG0 (i / i~e.0)
:ARG1 (p / possible-01~e.3
:ARG1 (a / add-02~e.4
:ARG0 (y / you~e.2)
:ARG1 (p2 / probation~e.5
:ARG1-of (c / contrast-01~e.7
:ARG2 (r / replace-01~e.12
:ARG1 p2
:ARG2~e.13 (t / time~e.15
:mod (j / jail~e.14))
:mod (j2 / just~e.10)))))))"""
amr_str1 = """(d1 / suppose-01~e.1
:ARG0 (i / i~e.0)
:ARG1 (p / possible-01~e.3
:ARG1 (a / add-02~e.4
:ARG0 (y / you~e.2)
:ARG1 (p2 / probation~e.5
:ARG1-of (c / contrast-01~e.7
:ARG2 (r / replace-01~e.12
:ARG1 p2
:mod (j2 / just~e.10)
:ARG2~e.13 (t / time~e.15
:mod (j / jail~e.14))
))))))"""
amr: AMR = AMR.parse_string(amr_str)
identified_concepts = IdentifiedConcepts()
identified_concepts.create_from_amr('amr_1', amr)
add_false_root(identified_concepts)
vector_of_parents = generate_parent_list_vector(amr, identified_concepts)
# transf parent vectors to vector of parents
# i s y p a p2 c j2 r j t
# 1 2 3 4 5 6 7 8 9 10 11
relations_dict = {
('suppose-01', 'i'): 'ARG0',
('suppose-01', 'possible-01'): 'ARG1',
('possible-01', 'add-02'): 'ARG1',
('add-02', 'you'): 'ARG0',
('add-02', 'probation'): 'ARG1',
('probation', 'contrast-01'): 'ARG1-of',
('contrast-01', 'replace-01'): 'ARG2',
('replace-01', 'probation'): 'ARG1',
('replace-01', 'time'): 'ARG2',
('replace-01', 'just'): 'mod',
('time', 'jail'): 'mod'
}
amr_node: Node = generate_amr_node_for_vector_of_parents(
identified_concepts, vector_of_parents, relations_dict)
generated_amr_str = amr_node.amr_print_with_reentrancy()
smatch = calculate_smatch(generated_amr_str, amr_str)
assert smatch == 1
def test_pre_and_post_processing_for_organization():
sentence = 'Some propaganda activities of ZF have soon become viewed as jokes by the people .'
amr_str = """(b / become-01~e.7
:ARG1 a
:ARG2 (v / view-02~e.8
:ARG0~e.11 (p2 / person~e.13)
:ARG1 (a / activity-06~e.2
:ARG0~e.3 (o / organization :wiki ZF
:name (n / name :op1 "ZF"~e.4))
:ARG1 (p / propaganda~e.1)
:quant (s / some~e.0))
:ARG2~e.9 (t / thing~e.10
:ARG2-of~e.10 (j / joke-01~e.10)))
:time~e.9 (s2 / soon~e.6))"""
amr: AMR = AMR.parse_string(amr_str)
amr, new_sentence, metadata = train_pre_processing(amr, sentence)
identified_concepts = IdentifiedConcepts()
identified_concepts.create_from_amr('', amr)
add_false_root(identified_concepts)
vector_of_parents = generate_parent_list_vector(amr, identified_concepts)
post_processing_on_parent_vector(identified_concepts, vector_of_parents,
new_sentence, metadata)
relations_dict = {
('become-01', 'activity-06'): 'ARG1',
('become-01', 'view-02'): 'ARG2',
('become-01', 'soon'): 'time',
('view-02', 'person'): 'ARG0',
('view-02', 'activity-06'): 'ARG1',
('view-02', 'thing'): 'ARG2',
('activity-06', 'organization'): 'ARG0',
('activity-06', 'propaganda'): 'ARG1',
('activity-06', 'some'): 'quant',
('organization', 'ZF'): 'wiki',
('organization', 'name'): 'name',
('name', 'ZF'): 'op1',
('thing', 'joke-01'): 'ARG2-of'
}
amr_node: Node = generate_amr_node_for_vector_of_parents(
identified_concepts, vector_of_parents, relations_dict)
generated_amr_str = amr_node.amr_print_with_reentrancy()
smatch = calculate_smatch(generated_amr_str, amr_str)
assert smatch == 1
def test_amr_print_with_literal():
r: Node = Node('realize-01')
neg: Node = Node('-')
i: Node = Node('i')
t2: Node = Node('threaten-01')
c2: Node = Node('country')
iran1: Node = Node(None, 'Iran')
n: Node = Node('name')
iran2: Node = Node(None, 'Iran')
h: Node = Node('huge')
s2: Node = Node('such')
c: Node = Node('cause-01')
a: Node = Node('amr-unknown')
s: Node = Node('simple')
r.add_child(neg, 'polarity')
r.add_child(i, 'ARG0')
r.add_child(t2, 'ARG1')
r.add_child(s, 'manner')
t2.add_child(c2, 'ARG0')
t2.add_child(h, 'mod')
t2.add_child(c, 'ARG1-of')
c2.add_child(iran1, 'wiki')
c2.add_child(n, 'name')
n.add_child(iran2, 'op1')
h.add_child(s2, 'degree')
c.add_child(a, 'ARG0')
generated_amr_str = r.amr_print()
expected_amr_str = """(r / realize-01 :polarity~e.3 -~e.3
:ARG0 (i / i~e.0)
:ARG1 (t2 / threaten-01~e.11
:ARG0 (c2 / country :wiki "Iran"
:name (n / name :op1 "Iran"~e.6))
:mod (h / huge~e.10
:degree (s2 / such~e.8))
:ARG1-of (c / cause-01~e.5
:ARG0~e.5 (a / amr-unknown~e.5)))
:manner (s / simple~e.1))"""
smatch = calculate_smatch(generated_amr_str,expected_amr_str)
assert smatch == 1
def test_generate_amr_node_for_vector_of_parents():
identified_concepts: IdentifiedConcepts = IdentifiedConcepts()
identified_concepts.ordered_concepts = [Concept('', 'ROOT'), # 0
Concept('i', 'it'), # 1
Concept('r', 'recommend-01'), # 2
Concept('v', 'vigorous'), # 3
Concept('a', 'advocate-01') # 4
]
predicted_vector_of_parents = [[-1], [4], [0], [4], [2]]
relations_dict = {
('recommend-01', 'advocate-01'): 'ARG1',
('advocate-01', 'it'): 'ARG1',
('advocate-01', 'vigorous'): 'manner'
}
amr: Node = generate_amr_node_for_vector_of_parents(identified_concepts,
predicted_vector_of_parents,
relations_dict)
generated_amr_str = amr.amr_print_with_reentrancy()
gold_amr_str = """(r / recommend-01~e.1
:ARG1 (a / advocate-01~e.4
:ARG1 (i / it~e.0)
:manner~e.2 (v / vigorous~e.3)))"""
smatch = calculate_smatch(generated_amr_str, gold_amr_str)
assert smatch == 1