def search_candidates(self, input_string):
editDistance = EditDistance()
for leaf in self.leaves:
for index in leaf.indices:
i = leaf.start
parent = leaf.parent
cand = True
p1 = -1
while cand and parent != None:
p1 = index - (i - parent.start)
p2 = index + (parent.end - i) + 1
if p1 < 0:
p1 = 0
if p2 > len(input_string):
p2 = len(input_string)
distance = editDistance.compute(input_string[p1:p2],
parent.pattern)
if distance <= parent.error:
parent = parent.parent
else:
p1 -= parent.error
counter = parent.error
withp1 = False
while counter != 0 and not withp1:
distance = editDistance.compute(
input_string[p1:p2], parent.pattern)
if distance <= parent.error:
parent = parent.parent
withp1 = True
else:
counter -= 1
p1 += 1
if not withp1:
p2 += parent.error
counter = parent.error
withp2 = False
while counter != 0 and not withp2:
distance = editDistance.compute(
input_string[p1:p2], parent.pattern)
if distance <= parent.error:
parent = parent.parent
withp2 = True
else:
counter -= 1
p2 -= 1
if not withp2:
cand = False
if cand:
self.indicesDict[p1] = input_string[p1:p2]
def edit_distance(self):
ed = EditDistance()
total_dist = 0
total_norm_dist = 0
op_count = {'m': 0, 'i': 0, 'd': 0, 'r': 0}
op_count_norm = {'m': 0, 'i': 0, 'd': 0, 'r': 0}
num_examples = len(self.gt)
num_examples = max(num_examples, 1)
for i in self.gt.keys():
gt = self.gt[i][0].split()
gen = self.gen[i][0].split()
max_len = float(max(len(gt), len(gen)))
max_len = max(max_len, 1.0)
dist = ed.compute(gt, gen)
total_dist += dist
total_norm_dist += dist / max_len
ops = ed.operations()
for op in ops:
op_count[op] += 1
op_count_norm[op] += 1.0 / max_len
mean_dist = total_dist / float(num_examples)
mean_norm_dist = total_norm_dist / float(num_examples)
for op in op_count:
op_count[op] /= float(num_examples)
op_count_norm[op] /= float(num_examples)
return mean_dist, mean_norm_dist, op_count, op_count_norm