s407259-mlworkshops/calculate_metrics.py
2019-05-16 12:33:45 +02:00

39 lines
951 B
Python

import pandas as pd
from jiwer import wer
df_data = pd.read_csv('./Infra/wikiniews_results.tsv', sep='\t', index_col=False, header=None,
skip_blank_lines=False, keep_default_na=False, names=[1,2,3,4,5])
correct_answers = 0
for index, row in df_data.iterrows():
wer_result = wer(row[2], row[3])
df_data.loc[index, 5] = wer_result
if wer_result == 0.0:
correct_answers += 1
wer_mean = df_data[5].mean()
print(wer_mean)
srr = (correct_answers*1)/len(df_data)
print(srr)
df_data.to_csv('./wikiniews_results_with_wer.tsv',sep='\t', header=None)
with open("wer_mean.txt", "w") as text_file:
text_file.write(str(df_data[5].mean()))
with open("srr.txt", "w") as text_file:
text_file.write(str(srr))
with open("historical_wer_mean.txt", "a") as text_file:
text_file.write(str(df_data[5].mean()) + '\n')
with open("historical_srr.txt", "a") as text_file:
text_file.write(str(srr) + '\n')