#!/usr/bin/env python3
import numpy as np
from utils import niceNameModel, PHNALL
from load import load_all_p1, load_all_p2
import pandas as pd
import matplotlib.pyplot as plt
if False:
dataP1 = load_all_p1(add_bleu=True)[[
'model', 'lang', 'fluency', 'adequacy', 'bleu'
]]
dataP1['mult'] = dataP1['fluency'] * dataP1['adequacy']
# data = data.drop(['fluency', 'adequacy'], axis=1)
for model in dataP1['model'].unique():
diff = dataP1[dataP1['model'] == model].groupby('lang').mean()
print(model)
print(diff)
print()
else:
dataP2 = load_all_p2(add_bleu=False)
def comp_occ(df, phnName):
return df[phnName].count() / df.shape[0]
def comp_sev(df, phnName):
tmp = df[phnName].dropna()
if tmp.shape[0] == 0:
return 0
else:
#!/usr/bin/env python3
from load import load_all_p1
from utils import niceNameDocArrow
import numpy as np
import re
import matplotlib.pyplot as plt
data = load_all_p1(add_bleu=True)
print('\n%%%' * 4)
for docName in sorted(
data['doc'].unique(),
key=lambda docName: data[data['doc'] == docName].mean()['mult'],
reverse=True):
dfDoc = data[data['doc'] == docName]
multAvg = dfDoc.mean()['mult']
errsAvg = dfDoc['errors'].mean()
bleuAvg = dfDoc['bleu'].mean()
bleuStd = np.sqrt(dfDoc['bleu'].std())
print(
f'{niceNameDocArrow(docName):<9} & {multAvg:10.2f} & {errsAvg:10.2f} & {bleuAvg:10.2f}',
'\\pmsmall{', f'{bleuStd:4.2f}', '} \\\\')
print('\\hline')
multAvg = data.mean()['mult']
errsAvg = data['errors'].mean()
bleuAvg = data['bleu'].mean()
bleuStd = np.sqrt(dfDoc['bleu'].std())
print(f'Average & {multAvg:10.2f} & {errsAvg:10.2f} & {bleuAvg:10.2f}',
#!/usr/bin/env python3
import numpy as np
from utils import niceNameModel, niceNamePhn, PHNALL
from load import load_all_p1, load_all_p2
import pandas as pd
pd.options.display.max_rows = 70
pd.options.display.max_colwidth = 50
data = load_all_p2(clear_badlines=True, add_bleu=True)
dataP1 = load_all_p1()
data['sev'] = data.apply(lambda row: sum(
[0 if np.isnan(row[phnName]) else row[phnName] for phnName in PHNALL]),
axis=1)