def process_row(row, writer):
actual_sentence = row['transcript']
target_sentence = row['target']
if len(actual_sentence.strip()) == 0:
return
if len(target_sentence.strip()) == 0:
return
print("{}-{}".format(actual_sentence, target_sentence))
diff = SentenceDiff(actual_sentence, target_sentence)
row['wer'] = diff.wer()
row['score'] = diff.chatterize_score() * 100
writer.writerow(row)
def test_substitutions(self):
list_of_lists = [["a", "b"], ["x", "y", "z"]]
result = SentenceDiff._all_substitutions(list_of_lists)
assert result == \
[("a","b"),
("b","a"),
("x", "y"),
("x", "z"),
("y", "x"),
("y", "z"),
("z", "x"),
("z", "y")]
def chatterize_score(self):
homonyms = SentenceDiff._homonyms(self.actual_lower)
max_similarity = -1
for homonym in homonyms:
similarity = self.similarity(homonym, self.target_lower)
if similarity > max_similarity:
max_similarity = similarity
pass_fail = "SUPER PASS" if max_similarity > THRESHOLD_SUPER_PASS \
else "PASS" if max_similarity > THRESHOLD_PASS \
else "FAIL"
return pass_fail, max_similarity
from sentence_diff import SentenceDiff
import csv
with open('edit_score_highlights.csv', newline='') as csv_file_in:
with open('edit_score__highlights_out.csv', 'w',
newline='') as csv_file_out:
writer = csv.writer(csv_file_out,
delimiter=',',
quotechar='"',
quoting=csv.QUOTE_MINIMAL)
writer.writerow([
"Actual Text", "Target Text", "Score By Words", "Score By Letters",
"New Score", "Attn", "New Score 2"
]) #"Updated Wer", "Wer Score"
reader = csv.reader(csv_file_in, delimiter=',', quotechar='"')
reader.__next__()
for row in reader:
actual_sentence = row[0]
target_sentence = row[1]
if len(actual_sentence.strip()) == 0:
continue
if len(target_sentence.strip()) == 0:
continue
print("{}-{}".format(actual_sentence, target_sentence))
diff = SentenceDiff(actual_sentence, target_sentence)
row.append(diff.chatterize_score() * 100)
writer.writerow(row)
def diff(actual_sentence, target_sentence):
return SentenceDiff(actual_sentence=actual_sentence,
target_sentence=target_sentence)
def normalize(self, text):
return \
SentenceDiff._remove_punctuation(
SentenceDiff._spell_out_numbers_in_word(
SentenceDiff._sound_out_dollars(
profanity.censor(text.lower(), 'x'))))
def __init__(self, actual, target):
SentenceDiff._assert_not_empty(actual, target)
self.actual = actual
self.target = target
self.actual_lower = self.normalize(actual)
self.target_lower = self.normalize(target)
def chatterize_score(actual_sentence, target_sentence):
diff = SentenceDiff(actual_sentence, target_sentence)
return diff.chatterize_score()
def test_normalize_1_dollar(self):
d = SentenceDiff("xx","xx")
assert d._normalize("here is $1 for you") == "here is 1 dollar for you"
def test_normalize_100_dollars(self):
d = SentenceDiff("xx","xx")
assert d._normalize("$100") == "100 dollars"
def test_sentence_homonymsdeserts(self):
result = SentenceDiff._homonyms("I love desert")
assert result == ["I love desert", "I love dessert"]