def match(self, verbose=False):
smatch.match_triple_dict.clear()
smatch.seed = self.seed
if self.seed is not None:
random.seed(seed)
gold_label = 'g'
test_label = 't'
gold_inst_orig, _, _ = self.gold_amr.parsed.get_triples()
test_inst_orig, _, _ = self.test_amr.parsed.get_triples()
self.gold_amr.parsed.rename_node(gold_label)
self.test_amr.parsed.rename_node(test_label)
gold_inst, gold_attr, gold_rel = self.gold_amr.parsed.get_triples()
test_inst, test_attr, test_rel = self.test_amr.parsed.get_triples()
gold_map = {a[1]: b[1] for a, b in zip(gold_inst, gold_inst_orig)}
test_map = {a[1]: b[1] for a, b in zip(test_inst, test_inst_orig)}
# best_mapping, self.best_match_num = smatch.get_best_match(gold_inst, gold_attr, gold_rel, test_inst, test_attr, test_rel, gold_label, test_label)
best_mapping, self.best_match_num = smatch.get_best_match(
test_inst, test_attr, test_rel, gold_inst, gold_attr, gold_rel,
test_label, gold_label)
gold_inst, gold_attr, gold_rel = self._rename(gold_inst, gold_attr,
gold_rel, gold_map)
test_inst, test_attr, test_rel = self._rename(test_inst, test_attr,
test_rel, test_map)
if verbose:
print('Gold instances:')
for instance in gold_inst:
print(' ', instance)
print('Test instances:')
for instance in test_inst:
print(' ', instance)
print("Best Match:",
smatch.print_alignment(best_mapping, gold_inst, test_inst),
file=sys.stderr)
# print("Matches:", self._matches(best_mapping, gold_inst, test_inst))
print("Matches:", self._matches(best_mapping, test_inst,
gold_inst))
gold_amr = self.gold_amr
test_amr = self.test_amr
gold_amr.instances, gold_amr.attributes, gold_amr.top = self._convert(
gold_inst, gold_attr)
gold_amr.relations = gold_rel
test_amr.instances, test_amr.attributes, test_amr.top = self._convert(
test_inst, test_attr)
test_amr.relations = test_rel
# gold_amr.matches, test_amr.matches = self._matches(best_mapping, gold_inst, test_inst)
test_amr.matches, gold_amr.matches = self._matches(
best_mapping, test_inst, gold_inst)
def score_amr_pair(ref_amr_line, rec_amr_line, restart_num, justinstance=False,
justattribute=False, justrelation=False):
# parse lines
amr1 = AMR.parse_AMR_line(ref_amr_line)
amr2 = AMR.parse_AMR_line(rec_amr_line)
# Fix prefix
prefix1 = "a"
prefix2 = "b"
# Rename node to "a1", "a2", .etc
amr1.rename_node(prefix1)
# Renaming node to "b1", "b2", .etc
amr2.rename_node(prefix2)
# get triples
(instance1, attributes1, relation1) = amr1.get_triples()
(instance2, attributes2, relation2) = amr2.get_triples()
# optionally turn off some of the node comparison
doinstance = doattribute = dorelation = True
if justinstance:
doattribute = dorelation = False
if justattribute:
doinstance = dorelation = False
if justrelation:
doinstance = doattribute = False
(best_mapping, best_match_num) = smatch.get_best_match(
instance1, attributes1, relation1,
instance2, attributes2, relation2,
prefix1, prefix2,
restart_num,
doinstance=doinstance,
doattribute=doattribute,
dorelation=dorelation
)
if justinstance:
test_triple_num = len(instance1)
gold_triple_num = len(instance2)
elif justattribute:
test_triple_num = len(attributes1)
gold_triple_num = len(attributes2)
elif justrelation:
test_triple_num = len(relation1)
gold_triple_num = len(relation2)
else:
test_triple_num = len(instance1) + len(attributes1) + len(relation1)
gold_triple_num = len(instance2) + len(attributes2) + len(relation2)
return best_match_num, test_triple_num, gold_triple_num
def match(self, verbose=False):
smatch.match_triple_dict.clear()
smatch.seed = self.seed
if self.seed is not None:
random.seed(seed)
gold_label = 'g'
test_label = 't'
gold_inst_orig,_,_ = self.gold_amr.parsed.get_triples()
test_inst_orig,_,_ = self.test_amr.parsed.get_triples()
self.gold_amr.parsed.rename_node(gold_label)
self.test_amr.parsed.rename_node(test_label)
gold_inst, gold_attr, gold_rel = self.gold_amr.parsed.get_triples()
test_inst, test_attr, test_rel = self.test_amr.parsed.get_triples()
gold_map = {a[1]:b[1] for a,b in zip(gold_inst, gold_inst_orig)}
test_map = {a[1]:b[1] for a,b in zip(test_inst, test_inst_orig)}
# best_mapping, self.best_match_num = smatch.get_best_match(gold_inst, gold_attr, gold_rel, test_inst, test_attr, test_rel, gold_label, test_label)
best_mapping, self.best_match_num = smatch.get_best_match(test_inst, test_attr, test_rel, gold_inst, gold_attr, gold_rel, test_label, gold_label)
gold_inst, gold_attr, gold_rel = self._rename(gold_inst, gold_attr, gold_rel, gold_map)
test_inst, test_attr, test_rel = self._rename(test_inst, test_attr, test_rel, test_map)
if verbose:
print('Gold instances:')
for instance in gold_inst:
print(' ', instance)
print('Test instances:')
for instance in test_inst:
print(' ', instance)
print("Best Match:", smatch.print_alignment(best_mapping, gold_inst, test_inst), file=sys.stderr)
# print("Matches:", self._matches(best_mapping, gold_inst, test_inst))
print("Matches:", self._matches(best_mapping, test_inst, gold_inst))
gold_amr = self.gold_amr
test_amr = self.test_amr
gold_amr.instances, gold_amr.attributes, gold_amr.top = self._convert(gold_inst, gold_attr)
gold_amr.relations = gold_rel
test_amr.instances, test_amr.attributes, test_amr.top = self._convert(test_inst, test_attr)
test_amr.relations = test_rel
# gold_amr.matches, test_amr.matches = self._matches(best_mapping, gold_inst, test_inst)
test_amr.matches, gold_amr.matches = self._matches(best_mapping, test_inst, gold_inst)
def match_amr(gold_amr, test_amr, verbose=False):
smatch.match_triple_dict.clear()
gold_label = 'g'
test_label = 't'
gold_inst_orig, _, _ = gold_amr.get_triples()
test_inst_orig, _, _ = test_amr.get_triples()
gold_amr.rename_node(gold_label)
test_amr.rename_node(test_label)
gold_inst, gold_attr, gold_rel = gold_amr.get_triples()
test_inst, test_attr, test_rel = test_amr.get_triples()
gold_map = {a[1]: b[1] for a, b in zip(gold_inst, gold_inst_orig)}
test_map = {a[1]: b[1] for a, b in zip(test_inst, test_inst_orig)}
# best_mapping, best_match_num = smatch.get_best_match(gold_inst, gold_attr, gold_rel, test_inst, test_attr, test_rel, gold_label, test_label)
best_mapping, best_match_num = smatch.get_best_match(
test_inst, test_attr, test_rel, gold_inst, gold_attr, gold_rel,
test_label, gold_label)
gold_inst, gold_attr, gold_rel = rename(gold_inst, gold_attr, gold_rel,
gold_map)
test_inst, test_attr, test_rel = rename(test_inst, test_attr, test_rel,
test_map)
if verbose:
print('Gold instances:')
for instance in gold_inst:
print(' ', instance)
print('Test instances:')
for instance in test_inst:
print(' ', instance)
print("Best Match:",
smatch.print_alignment(best_mapping, gold_inst, test_inst),
file=sys.stderr)
# print("Matches:", matches(best_mapping, gold_inst, test_inst))
print("Matches:", matches(best_mapping, test_inst, gold_inst))
amr1 = convert(gold_amr, gold_inst, gold_attr, gold_rel)
amr2 = convert(test_amr, test_inst, test_attr, test_rel)
# amr1.matches, amr2.matches = matches(best_mapping, gold_inst, test_inst)
amr2.matches, amr1.matches = matches(best_mapping, test_inst,
gold_inst)
return amr1, amr2, best_match_num
def match_pair_2(pair):
(lst_amr1, dic_amr1), (lst_amr2, dic_amr2) = pair
prefix1 = 'a'
prefix2 = 'b'
smatch.match_triple_dict.clear()
amr1 = parse_relations(lst_amr1, dic_amr1)
amr2 = parse_relations(lst_amr2, dic_amr2)
amr1.rename_node(prefix1) # Rename node to "a1", "a2", .etc
amr2.rename_node(prefix2) # Rename node to "b1", "b2", .etc
inst1, attrib1, rel1 = amr1.get_triples()
inst2, attrib2, rel2 = amr2.get_triples()
_, best_match_num = smatch.get_best_match(inst1, attrib1, rel1, inst2,
attrib2, rel2, prefix1, prefix2)
num_test = len(inst1) + len(attrib1) + len(rel1)
num_gold = len(inst2) + len(attrib2) + len(rel2)
return best_match_num, num_test, num_gold
def match_amr(gold_amr, test_amr, verbose=False):
smatch.match_triple_dict.clear()
gold_label = 'g'
test_label = 't'
gold_inst_orig,_,_ = gold_amr.get_triples()
test_inst_orig,_,_ = test_amr.get_triples()
gold_amr.rename_node(gold_label)
test_amr.rename_node(test_label)
gold_inst, gold_attr, gold_rel = gold_amr.get_triples()
test_inst, test_attr, test_rel = test_amr.get_triples()
gold_map = {a[1]:b[1] for a,b in zip(gold_inst, gold_inst_orig)}
test_map = {a[1]:b[1] for a,b in zip(test_inst, test_inst_orig)}
# best_mapping, best_match_num = smatch.get_best_match(gold_inst, gold_attr, gold_rel, test_inst, test_attr, test_rel, gold_label, test_label)
best_mapping, best_match_num = smatch.get_best_match(test_inst, test_attr, test_rel, gold_inst, gold_attr, gold_rel, test_label, gold_label)
gold_inst, gold_attr, gold_rel = rename(gold_inst, gold_attr, gold_rel, gold_map)
test_inst, test_attr, test_rel = rename(test_inst, test_attr, test_rel, test_map)
if verbose:
print('Gold instances:')
for instance in gold_inst:
print(' ', instance)
print('Test instances:')
for instance in test_inst:
print(' ', instance)
print("Best Match:", smatch.print_alignment(best_mapping, gold_inst, test_inst), file=sys.stderr)
# print("Matches:", matches(best_mapping, gold_inst, test_inst))
print("Matches:", matches(best_mapping, test_inst, gold_inst))
amr1 = convert(gold_amr, gold_inst, gold_attr, gold_rel)
amr2 = convert(test_amr, test_inst, test_attr, test_rel)
# amr1.matches, amr2.matches = matches(best_mapping, gold_inst, test_inst)
amr2.matches, amr1.matches = matches(best_mapping, test_inst, gold_inst)
return amr1, amr2, best_match_num
def compute_files(user1, user2, file_list, dir_pre, start_num):
"""
Compute the smatch scores for a file list between two users
Args:
user1: user 1 name
user2: user 2 name
file_list: file list
dir_pre: the file location prefix
start_num: the number of restarts in smatch
Returns:
smatch f score.
"""
match_total = 0
test_total = 0
gold_total = 0
for fi in file_list:
file1 = dir_pre + user1 + "/" + fi + ".txt"
file2 = dir_pre + user2 + "/" + fi + ".txt"
if not os.path.exists(file1):
print("*********Error: ",
file1,
"does not exist*********",
file=ERROR_LOG)
return -1.00
if not os.path.exists(file2):
print("*********Error: ",
file2,
"does not exist*********",
file=ERROR_LOG)
return -1.00
try:
file1_h = open(file1, "r")
file2_h = open(file2, "r")
except IOError:
print("Cannot open the files", file1, file2, file=ERROR_LOG)
break
cur_amr1 = smatch.get_amr_line(file1_h)
cur_amr2 = smatch.get_amr_line(file2_h)
if cur_amr1 == "":
print("AMR 1 is empty", file=ERROR_LOG)
continue
if cur_amr2 == "":
print("AMR 2 is empty", file=ERROR_LOG)
continue
amr1 = amr.AMR.parse_AMR_line(cur_amr1)
amr2 = amr.AMR.parse_AMR_line(cur_amr2)
test_label = "a"
gold_label = "b"
amr1.rename_node(test_label)
amr2.rename_node(gold_label)
(test_inst, test_rel1, test_rel2) = amr1.get_triples()
(gold_inst, gold_rel1, gold_rel2) = amr2.get_triples()
if verbose:
print("Instance triples of file 1:",
len(test_inst),
file=DEBUG_LOG)
print(test_inst, file=DEBUG_LOG)
print("Attribute triples of file 1:",
len(test_rel1),
file=DEBUG_LOG)
print(test_rel1, file=DEBUG_LOG)
print("Relation triples of file 1:",
len(test_rel2),
file=DEBUG_LOG)
print(test_rel2, file=DEBUG_LOG)
print("Instance triples of file 2:",
len(gold_inst),
file=DEBUG_LOG)
print(gold_inst, file=DEBUG_LOG)
print("Attribute triples of file 2:",
len(gold_rel1),
file=DEBUG_LOG)
print(gold_rel1, file=DEBUG_LOG)
print("Relation triples of file 2:",
len(gold_rel2),
file=DEBUG_LOG)
print(gold_rel2, file=DEBUG_LOG)
(best_match, best_match_num) = smatch.get_best_match(
test_inst, test_rel1, test_rel2, gold_inst, gold_rel1, gold_rel2,
test_label, gold_label)
if verbose:
print("best match number", best_match_num, file=DEBUG_LOG)
print("Best Match:",
smatch.print_alignment(best_match, test_inst, gold_inst),
file=DEBUG_LOG)
match_total += best_match_num
test_total += (len(test_inst) + len(test_rel1) + len(test_rel2))
gold_total += (len(gold_inst) + len(gold_rel1) + len(gold_rel2))
smatch.match_triple_dict.clear()
(precision, recall, f_score) = smatch.compute_f(match_total, test_total,
gold_total)
return "%.2f" % f_score
def compute_files(user1, user2, file_list, dir_pre, start_num):
"""
Compute the smatch scores for a file list between two users
Args:
user1: user 1 name
user2: user 2 name
file_list: file list
dir_pre: the file location prefix
start_num: the number of restarts in smatch
Returns:
smatch f score.
"""
match_total = 0
test_total = 0
gold_total = 0
for fi in file_list:
file1 = dir_pre + user1 + "/" + fi + ".txt"
file2 = dir_pre + user2 + "/" + fi + ".txt"
if not os.path.exists(file1):
print >> ERROR_LOG, "*********Error: ", file1, "does not exist*********"
return -1.00
if not os.path.exists(file2):
print >> ERROR_LOG, "*********Error: ", file2, "does not exist*********"
return -1.00
try:
file1_h = open(file1, "r")
file2_h = open(file2, "r")
except IOError:
print >> ERROR_LOG, "Cannot open the files", file1, file2
break
cur_amr1 = smatch.get_amr_line(file1_h)
cur_amr2 = smatch.get_amr_line(file2_h)
if cur_amr1 == "":
print >> ERROR_LOG, "AMR 1 is empty"
continue
if cur_amr2 == "":
print >> ERROR_LOG, "AMR 2 is empty"
continue
amr1 = amr.AMR.parse_AMR_line(cur_amr1)
amr2 = amr.AMR.parse_AMR_line(cur_amr2)
test_label = "a"
gold_label = "b"
amr1.rename_node(test_label)
amr2.rename_node(gold_label)
(test_inst, test_rel1, test_rel2) = amr1.get_triples()
(gold_inst, gold_rel1, gold_rel2) = amr2.get_triples()
if verbose:
print >> DEBUG_LOG, "Instance triples of file 1:", len(test_inst)
print >> DEBUG_LOG, test_inst
print >> DEBUG_LOG, "Attribute triples of file 1:", len(test_rel1)
print >> DEBUG_LOG, test_rel1
print >> DEBUG_LOG, "Relation triples of file 1:", len(test_rel2)
print >> DEBUG_LOG, test_rel2
print >> DEBUG_LOG, "Instance triples of file 2:", len(gold_inst)
print >> DEBUG_LOG, gold_inst
print >> DEBUG_LOG, "Attribute triples of file 2:", len(gold_rel1)
print >> DEBUG_LOG, gold_rel1
print >> DEBUG_LOG, "Relation triples of file 2:", len(gold_rel2)
print >> DEBUG_LOG, gold_rel2
(best_match, best_match_num) = smatch.get_best_match(test_inst, test_rel1, test_rel2,
gold_inst, gold_rel1, gold_rel2,
test_label, gold_label)
if verbose:
print >> DEBUG_LOG, "best match number", best_match_num
print >> DEBUG_LOG, "Best Match:", smatch.print_alignment(best_match, test_inst, gold_inst)
match_total += best_match_num
test_total += (len(test_inst) + len(test_rel1) + len(test_rel2))
gold_total += (len(gold_inst) + len(gold_rel1) + len(gold_rel2))
smatch.match_triple_dict.clear()
(precision, recall, f_score) = smatch.compute_f(match_total, test_total, gold_total)
return "%.2f" % f_score