-
Notifications
You must be signed in to change notification settings - Fork 0
/
evaluate.py
325 lines (231 loc) · 9.42 KB
/
evaluate.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
from GaelicTextNormaliser import TextNormaliser
from utils.utils import fetch_text
from utils.preprocess import preprocess_doc
import difflib, re
from tqdm.auto import tqdm as tt
all_apostrophes = re.compile(r"[’'`’‘]")
def count_Er_toks(gold_toks, bef_toks, norm_toks):
errors = {'correctBefore' : 0,
'correctAfter' : 0,
'incorrectBefore' : 0,
'incorrectAfter' : 0}
for i in gold_toks:
if i not in bef_toks:
errors['incorrectBefore']+=1
if i in bef_toks:
errors['correctBefore']+=1
if i in norm_toks:
errors['correctAfter']+=1
for i in norm_toks:
if i not in gold_toks and i not in bef_toks:
errors['incorrectAfter'] += 1
return errors
def calc_ER(errors):
try:
return (errors['correctAfter']-errors['correctBefore'])/errors['incorrectBefore']
except ZeroDivisionError:
return 0
def ER_of_line(b_line, g_line, n_line):
g_toks=g_line.split(' ')
b_toks=b_line.split(' ')
n_toks=n_line.split(' ')
errors=count_Er_toks(g_toks,b_toks,n_toks)
return calc_ER(errors)
def count_tokens(tokens):
counts = {}
for token in tokens:
if token not in counts:
counts[token] = 1
else:
counts[token] += 1
return counts
def calc_accuracy(t_line, g_line):
t_toks=t_line.split(' ')
g_toks=g_line.split(' ')
t_counts = count_tokens(t_toks)
g_counts = count_tokens(g_toks)
not_in = 0
for tok in t_toks:
if tok in g_counts:
t_counts[tok] -= 1
else:
not_in += 1
return 1 - (sum(t_counts.values()) / sum(g_counts.values()))
def evaluate_text(before, gold, normalised):
b_accuracy = []
a_accuracy = []
ers = []
for b_line, g_line, n_line in zip(before, gold, normalised):
bef_accuracy = calc_accuracy(b_line, g_line)
b_accuracy.append(bef_accuracy)
af_accuracy = calc_accuracy(n_line, g_line)
a_accuracy.append(af_accuracy)
er = ER_of_line(b_line, g_line, n_line)
if er >= 0:
ers.append(er)
else:
ers.append(0)
#print(f'Bef: {bef_accuracy}\nAf: {af_accuracy}\nEr: {er}\n')
avg_bef = sum(b_accuracy)/len(b_accuracy)
avg_af = sum(a_accuracy)/ len(a_accuracy)
avg_ers = sum(ers)/len(ers)
print(f'Avg Bef: {avg_bef}\nAvg Af: {avg_af}\nAvg ER: {avg_ers}')
return b_accuracy, a_accuracy, ers
def calc_confusion(tp_common_words, fn_added_words, fp_removed_words, tok=None, verbose=None):
precision_ppv = len(tp_common_words) / (len(tp_common_words)+len(fp_removed_words))
recall_TPR = len(tp_common_words) / (len(tp_common_words)+len(fn_added_words))
try:
f1_score = 2 * ((precision_ppv * recall_TPR)/ (precision_ppv+recall_TPR))
except:
f1_score = 0
if verbose:
print(f'Precision: {precision_ppv}')
print(f'Recall: {recall_TPR}')
print(f'F1 Score: {f1_score}')
return precision_ppv, recall_TPR, f1_score
def gather_errors(doc_a, doc_b, tok=None):
doc_a_toks = doc_a.split(' ')
doc_b_toks = doc_b.split(' ')
differences = difflib.ndiff(doc_a_toks, doc_b_toks)
tp_common_words = []
fn_added_words = []
fp_removed_words = []
num=None
for i in differences:
temp = [i for i in i.split(' ') if i not in ['', '?', '^\n', '^^^\n']]
if tok:
try:
if temp[1] == tok:
num=True
#print(' '.join([i for i in differences]))
except:
None
if len(temp) < 1:
continue
if temp[0] == '+':
try:
fn_added_words.append(temp[1])
except:
continue
print(temp)
elif temp[0] == '-':
try:
fp_removed_words.append(temp[1])
except:
continue
else:
tp_common_words.append(temp[0])
return tp_common_words, fn_added_words, fp_removed_words, num
def update_wordlist(wordlist, fn, fp, line_idx):
for words in fn:
if words not in wordlist:
wordlist[words] = {'count' : 1,
'idxs' : [line_idx]}
else:
wordlist[words]['count']+=1
wordlist[words]['idxs'].append(line_idx)
for words in fp:
if words not in wordlist:
wordlist[words] = {'count' : 1,
'idxs' : [line_idx]}
else:
wordlist[words]['count']+=1
wordlist[words]['idxs'].append(line_idx)
return wordlist
def calc_metrics(input_text, gold_text, norm_text, bad_words,idx=None, verbose=None, tok=None):
metrics = {}
bef_tp_common_words, bef_fn_added_words, bef_fp_removed_words, return_line = gather_errors(input_text, gold_text, tok=tok)
norm_tp_common_words, norm_fn_added_words, norm_fp_removed_words, _ = gather_errors(norm_text, gold_text, tok=None)
incorrect_before_normalisation = len(bef_fp_removed_words)
correct_before_normalisation = len(bef_tp_common_words)
correct_after_normalisation = len(norm_tp_common_words)
incorrect_after_normalisation = len(norm_fp_removed_words)
bef_accuracy = len(bef_tp_common_words) / len(gold_text.split(' '))
norm_accuracy = len(norm_tp_common_words) / len(gold_text.split(' '))
#error_reduction = (correct_after_normalisation-correct_before_normalisation)/incorrect_before_normalisation
cb = (len(bef_tp_common_words)/len(gold_text.split(' ')))
ib = (len(bef_fp_removed_words)+len(bef_fn_added_words)/len(gold_text.split(' ')))
ca = (correct_after_normalisation/len(gold_text.split(' ')))
bad_words = update_wordlist(bad_words, norm_fn_added_words, norm_fp_removed_words, idx)
# before normalisation
bef_precision_ppv, bef_recall_TPR, bef_f1_score =calc_confusion(bef_tp_common_words, bef_fn_added_words, bef_fp_removed_words)
# normalised confusion
norm_precision_ppv, norm_recall_TPR, norm_f1_score =calc_confusion(norm_tp_common_words, norm_fn_added_words, norm_fp_removed_words)
try:
er = ((ca-cb)/ib)
except ZeroDivisionError:
er = 1
#print(f'Error 2: {er}\n')
if er < 0:
er = 0
if verbose:
print(f'{len(norm_tp_common_words)} True Positives')
print(f'{len(norm_fn_added_words)} False Negatives')
print(f'{len(norm_fp_removed_words)} False Positives\n')
print(f'\nError Reduction: {er}')
print(f'Accuracy Before Normalisation: {bef_accuracy}')
print(f'Accuracy After Normalisation: {norm_accuracy}')
metrics['bef_accuracy']=bef_accuracy
metrics['af_accuracy']=norm_accuracy
metrics['bef_precision']=bef_precision_ppv
metrics['bef_recall']=bef_recall_TPR
metrics['bef_f1']=bef_f1_score
metrics['norm_precision']=norm_precision_ppv
metrics['norm_recall']=norm_recall_TPR
metrics['norm_f1']=norm_f1_score
return metrics, bad_words, return_line
def update_metrics(upto, metrics, updated):
for i in updated:
metrics[i]+=(updated[i]/upto)
return metrics
def gather_metrics(normaliser, target_text, gold_text):
num=0
bad_words={}
total_bef_accuracy=[]
total_af_accuracy = []
total_er=[]
precision=[]
recall=[]
total_f1_score=[]
return_line = None
error_lines=[]
for bef, af in tt(zip(target_text, gold_text)):
bef = re.sub(all_apostrophes, "'", bef)
af = re.sub(all_apostrophes, "'", af)
normalised_text = normaliser.normalise_doc(doc=bef)
normalised_text = re.sub(all_apostrophes, "'", normalised_text)
updated_metrics, bad_words, return_line = calc_metrics(bef, af, normalised_text, bad_words,idx=num, tok="a'")
if return_line==True:
error_lines.append(num)
if num==0:
metrics={i:0 for i in updated_metrics}
metrics=update_metrics(1, metrics, updated_metrics)
if num == 1:
break
num+=1
return metrics, bad_words,
def print_metrics(metrics):
for key, val in metrics.items():
val = val
print(f'{key}: {val}\n')
if metrics['af_accuracy'] > metrics['bef_accuracy']:
improvement = round((metrics['af_accuracy']-metrics['bef_accuracy'])*100,7)
print('Improved accuracy of {}%'.format(improvement))
def fetch_text(path):
with open(path) as file:
doc = file.read()
doc = re.sub('\n', ' ', doc)
return doc
def evaluate(bef_path, af_path, normaliser):
bef_doc = fetch_text(bef_path)
af_doc = fetch_text(af_path)
metrics, bad_words = gather_metrics(normaliser, [bef_doc], [af_doc])
print_metrics(metrics)
return metrics
if __name__ == '__main__':
norm=TextNormaliser(from_config='config.yaml')
tad_before = fetch_text('resources/data/test_cases/pre_goc.txt')
tad_after = fetch_text('resources/data/test_cases/goc.txt')
#normalised = norm.normalise_doc(doc_path='resources/data/test_cases/pre_goc.txt').split('\n')
#b_accuracy, a_accuracy, ers = evaluate_text(tad_before, tad_after, normalised)
metrics, bad_words = gather_metrics([norm], [tad_before], [tad_after])