def evaluate_predictions(**kwargs):
distance = 0
diff_count = 0
for row_i, row in search_utils.chinese_names.iterrows():
# if row_i > 10:
# break
if row_i % 50 == 0:
print("{}% complete".format(100 * row_i /
search_utils.chinese_names.shape[0]))
english, _, _, target_pinyin, _, _, _ = row
english = search_utils.normalize(english)
target_pinyin = ''.join(
filter(lambda x: x != ' ', search_utils.normalize(target_pinyin)))
#print(english)
if search_utils.is_vowel(english[0]): english = 'S' + english
if search_utils.is_vowel(english[-1]): english = english + 'E'
output_pinyin = ''.join(find_closest_pinyin(english, **kwargs)[1])
if output_pinyin != target_pinyin:
#print (english, output_pinyin, target_pinyin)
diff_count += 1
distance += edit_distance.edit_distance_pinyin(
target_pinyin, output_pinyin)
print(
"Out of {} names, {} were different, with an average edit distance of {} ({} for just the different pairs)"
.format(search_utils.chinese_names.shape[0], diff_count,
distance / search_utils.chinese_names.shape[0],
distance / diff_count))
def evaluate_predictions():
all_names = search_utils.chinese_names
distance = 0
diff_count = 0
for row_i, row in all_names.iterrows():
if row_i % 20 == 0:
print("{}% complete".format(100 * row_i / all_names.shape[0]))
english, _, _, target_pinyin, _, _, _ = row
result = baseline(english)
result_pinyin = pinyin(result)
target_pinyin = ''.join(
filter(lambda x: x != ' ', search_utils.normalize(target_pinyin)))
output_pinyin = ''.join([seg[0] for seg in result_pinyin])
if output_pinyin != target_pinyin:
#print (english, output_pinyin, target_pinyin)
diff_count += 1
distance += edit_distance.edit_distance_pinyin(
target_pinyin, output_pinyin)
print(
"Out of {} names, {} were different, with an average edit distance of {} ({} for just the different pairs)"
.format(all_names.shape[0], diff_count, distance / all_names.shape[0],
distance / diff_count))
def process_baseline(oracle_csv):
df = pd.ExcelFile(oracle_csv).parse('Sheet1')
df_baseline_pinyin = pd.ExcelFile(os.path.join("..", "data", "proposal", "BaselineResponses.xlsx")).parse('Sheet1')
#df = pd.read_csv(oracle_csv)
names = df["English"]
o1 = df["Pinyin_O1"]
o2 = df["Pinyin_O2"]
bp = df_baseline_pinyin["Baseline"]
distance = 0
diff_count = 0
for name, name1, name2, pinyin in zip(names, o1, o2, bp):
if name != name1 or name != name2:
diff_count += 1
baseline_guess = baseline.baseline(name)
print(baseline_guess)
dist_o1 = edit_distance.edit_distance_pinyin(pinyin, name1)
print("Distance between", pinyin, "and", name1, ":", dist_o1)
dist_o2 = edit_distance.edit_distance_pinyin(pinyin, name2)
print("Distance between", pinyin, "and", name2, ":", dist_o2)
distance += ((dist_o1 + dist_o2) / 2)
# take the average over ALL names
return (distance/len(names), diff_count, len(names))
def process_oracle(oracle_csv):
df = pd.ExcelFile(oracle_csv).parse('Sheet1')
#df = pd.read_csv(oracle_csv)
o1 = df["Pinyin_O1"]
o2 = df["Pinyin_O2"]
distance = 0
diff_count = 0
for name1, name2 in zip(o1, o2):
if name1 != name2:
diff_count += 1
dist = edit_distance.edit_distance_pinyin(name1, name2) # 1. penalizes wrong tones as 0.5
# 2. doesn't do the "count characters in common" thing because these are longer
# only print stuff that's different
print("Distance between", name1, "and", name2, ":", dist)
distance += dist
# take the average over ALL names, not just the ones that were wrong
return (distance/len(o1), diff_count, len(o1))
def evaluateAllLines(encoder, decoder):
distance = 0
diff_count = 0
for pair in pairs:
output_name = evaluate(encoder, decoder, pair[0])
# remove the space for edit distance calculations for consistency with baseline
output_name = ''.join(filter(
lambda l: l != ' ',
output_name[:-1])) # need to get rid of the <EOS> string at end
target_name = ''.join(filter(lambda l: l != ' ', pair[1]))
if output_name != target_name:
#print(output_name, target_name)
diff_count += 1
distance += edit_distance.edit_distance_pinyin(
target_name, output_name)
print(
"Out of {} names, {} were different, with an average edit distance of {} ({} for just the different pairs)"
.format(len(pairs), diff_count, distance / len(pairs),
distance / diff_count))