-
Notifications
You must be signed in to change notification settings - Fork 1
/
util.py
567 lines (492 loc) · 15.1 KB
/
util.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
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
from math import log, sqrt, floor
from collections import Counter
from statistics import mean
from string import punctuation
import operator, zipfile, os
from datasets import load_dataset
from datasets.utils.logging import disable_progress_bar
from sklearn.feature_extraction.text import CountVectorizer
import pandas as pd
import smtplib
disable_progress_bar()
def load_huggingface(dataset_name, subset, split, column_name):
"""
Load dataset from the HuggingFace datasets library.
Arguments:
dataset: dataset name (string),
subset: data subset (string),
split: data split (string),
column_name: name of the column to analyze (string)
Returns:
texts: list of strings,
infiles: doc indices
"""
# Load dataset
if subset.strip() and split.strip():
dataset = load_dataset(dataset_name, subset, split=split)
elif subset.strip() and not split.strip():
dataset = load_dataset(dataset_name, subset)
elif not subset.strip() and split.strip():
dataset = load_dataset(dataset_name, split=split)
else: # not subset.strip() and not split.strip()
dataset = load_dataset(dataset_name)
# Convert to DataFrame
df = pd.DataFrame(dataset)
# prepare output
texts = df[column_name].tolist()
infiles = df.index.tolist()
return texts, infiles
def load_data(input_format, input_dir, text_column=None, delimiter=None):
"""
Load the dataset.
Arguments:
input_format: input format specified in the config file (csv or zip),
input_dir: input directory specified in the config file,
text_column: if input_format==csv, column name containing texts,
delimiter: if input_format==csv, delimiter for reading the csv file.
Returns:
texts: list of strings,
infiles: doc indices
"""
if input_format == 'zip': # zip folder with txt
filenames = []
texts = []
with zipfile.ZipFile(input_dir, 'r') as zip_file:
for file_info in zip_file.infolist():
if file_info.filename.endswith('.txt'):
filename = os.path.basename(file_info.filename)
with zip_file.open(file_info) as txt_file:
text = txt_file.read().decode('utf-8') # Assuming UTF-8 encoding
texts.append(text)
filenames.append(filename)
df = pd.DataFrame(data={'filename': filenames, 'text': texts})
df = df.sort_values('filename')
texts = list(df.text)
infiles = list(df.filename)
elif input_format == 'csv':
df = pd.read_csv(input_dir, delimiter=delimiter)
texts = list(df[text_column])
infiles = list(df.index)
else: # directory of txt files
raise ValueError("Input type must be 'csv' or 'zip'.")
return texts, infiles
#BASELINE_SYLLABIFIER________________________________________________________________
def get_n_syllables(word, dic):
"""
Syllabification based on Pyphen library.
Arguments:
word: str
dic: pyphen instance
Returns:
n_syllables: int
"""
result = dic.inserted(word).split('-')
return len(result)
#STATISTICS__________________________________________________________________________
# def contains_negation(tokenized_sentence, negators={'niet', 'niets', 'geen', 'nooit', 'niemand', 'nergens', 'noch'}):
# """
# Detects negation in a sentence
# Arguments
# tokenized_sentence: list of tokens
# negators: lexicon indicating negation
# Returns
# Bool
# """
# for n in negators:
# if n in set(t.lower() for t in tokenized_sentence):
# return True
# return False
def ratio_content_words(doc):
"""
Computes ratio of content words to all words (PUNCT, SYM, and X are excluded in this computation)
Arguments:
doc: Spacy doc object
Returns
Ratio of content words
"""
content = [t.text for s in doc.sents for t in s if t.pos_ in {'ADJ', 'ADV', 'NOUN', 'VERB', 'PROPN'}]
funct = [t.text for s in doc.sents for t in s if t.pos_ not in {'ADP', 'AUX', 'CCONJ', 'DET', 'NUM', 'PART', 'PRON', 'SCONJ'}]
try:
return len(content)/(len(content)+len(funct))
except (ValueError, ZeroDivisionError):
return 0
def get_passive_ratio(doc, matcher):
"""
Computes ratio of sentences that contain a passive verb construction.
Arguments:
doc: Spacy doc object
matcher: Spacy matcher object
Returns:
Ratio of passive sentencs
"""
total = 0
passive = 0
for s in doc.sents:
total += 1
matches = matcher(s)
if matches:
passive += 1
return round(passive/total, 3)
#LEXICAL_RICHNESS_SCORES_____________________________________________________________
"""
Various functions for computing lexical richness scores
"""
def ttr(n_types,n_tokens):
"""
Computes type-token ratio.
Input:
n_types: number of unique words
n_tokens: number of words
Returns:
TTR
"""
return round(n_types/n_tokens, 3)
def sttr(tokens, span_size):
"""
Computes standardized type-token ratio (per 100 tokens).
Input:
tokens: list of tokens
span_size: size of the segments on which TTR is computed.
Returns:
STTR if len(tokens) > span_size, else TTR
"""
if len(tokens) < span_size:
n_tokens = len(tokens)
n_types = len(set(tokens))
return round(n_types/n_tokens, 3)
ttr_scores = []
for i in range(0, len(tokens), span_size):
segment = tokens[i:i+span_size]
n_types = len(set(segment))
n_tokens = len(segment)
ttr_scores.append(ttr(n_types, n_tokens))
return round(mean(ttr_scores), 3)
"""
Following functions are variations on TTR,
but all based on number of types and tokens.
"""
def rttr(n_types,n_tokens):
return round(n_types/sqrt(n_tokens), 3)
def cttr(n_types,n_tokens):
return round(n_types/sqrt(2*n_tokens), 3)
def Herdan(n_types,n_tokens):
try:
return round(log(n_types)/log(n_tokens), 3)
except (ValueError, ZeroDivisionError):
return None
def Summer(n_types,n_tokens):
try:
return round(log(log(n_types))/log(log(n_tokens)), 3)
except (ValueError, ZeroDivisionError):
return None
def Dugast(n_types,n_tokens):
try:
return round((log(n_tokens)**2)/(log(n_tokens)-log(n_types)), 3)
except (ValueError, ZeroDivisionError):
return None
def Maas(n_types,n_tokens):
try:
return round((log(n_tokens)-log(n_types))/(log(n_tokens)**2), 3)
except (ValueError, ZeroDivisionError):
return None
#READABILITY SCORES__________________________________________________________________
"""
Various functions for computing readability scores
"""
def ARI(n_char, n_tokens, n_sentences):
return round(4.71*(n_char/n_tokens)+0.5*(n_tokens/n_sentences)-21.43, 3)
def ColemanLiau(tokens, tokenized_sentences):
if len(tokens) < 100:
return None
chunks = [tokens[i:i+100] for i in range(0, len(tokens), 100) if i+100<=len(tokens)]
L = mean([len(''.join(chunk)) for chunk in chunks]) # avg. n char per 100 tokens
S = len(tokenized_sentences)/len(tokens)*100 # avg. n sent per 100 tokens
return round(0.0588*L-0.296*S-15.8, 3)
def Flesch(ASL, ASW):
return round(206.835-(1.015*ASL)-(84.6*ASW), 3)
def Fog(ASL, syllables):
syllables = [s for sent in syllables for s in sent]
if len(syllables) < 100:
return None
PHW = len([s for s in syllables if s >= 3])/len(syllables) # percentage hard words, i.e. at least three syllables
return round(0.4*(ASL + PHW), 3)
def Kincaid(ASL, ASW):
return round((0.39*ASL)+(11.8*ASW)-15.59, 3)
def LIX(n_tokens, n_sentences, n_long_tokens):
return round((n_tokens/n_sentences)+(n_long_tokens*100/n_tokens), 3)
def RIX(n_long_tokens, n_sentences):
return round(n_long_tokens/n_sentences, 3)
def SMOG(sample):
length = len(sample)
if length < 30:
return None
else:
i = floor(length/3)
sample = sample[:10] + sample[i:i+10] + sample[-10:] # select first 10 sentences, last 10 sentences, and 10 sentences in the middle
sample = [s for sent in sample for s in sent]
n_polysyllabic = len([s for s in sample if s > 2]) # check if more than 2 syllables
return round(sqrt(n_polysyllabic) + 3, 3)
def interpret_readability(score, name):
"""
Converts readability score to interpretable results.
Arguments:
score: float
name: readability metric name
Returns:
string (or None if score is None)
"""
if not score:
return None
elif name == 'Flesch reading ease':
if score >= 90:
return "USA 5th Grade"
elif score >= 80:
return "USA 6th Grade"
elif score >= 70:
return "USA 7th Grade"
elif score >= 60:
return "USA 8th-9th Grade"
elif score >= 50:
return "USA 10th-12th Grade"
elif score >= 30:
return "USA College Student"
elif score >= 10:
return "USA College Graduate"
else:
return "Professional"
if name in {'ARI', 'Flesch-Kincaid Grade Level', 'Coleman-Liau', 'Gunning Fog', 'SMOG'}:
if score < 1:
return "USA Kindergarten"
elif score < 2:
return "USA 1st Grade"
elif score < 3:
return "USA 2nd Grade"
elif score < 4:
return "USA 3rd Grade"
elif score < 5:
return "USA 4th Grade"
elif score < 6:
return "USA 5th Grade"
elif score < 7:
return "USA 6th Grade"
elif score < 8:
return "USA 7th Grade"
elif score < 9:
return "USA 8th Grade"
elif score < 10:
return "USA 9th Grade"
elif score < 11:
return "USA 10th Grade"
elif score < 12:
return "USA 11th Grade"
elif score < 13:
return "USA 12th Grade"
elif score < 14:
return "USA College Freshman"
elif score < 15:
return "USA College Sophomore"
elif score < 16:
return "USA College Junior"
elif score < 17:
return "USA College Senior"
else:
return "USA College Graduate"
elif name == 'LIX':
if score < 30:
return "Very easy"
elif score < 40:
return "Easy"
elif score < 50:
return "Medium"
elif score < 60:
return "Difficult"
else:
return "Very difficult"
elif name == 'RIX':
if score < 0.2:
return "USA 1st Grade"
elif score < 0.5:
return "USA 2nd Grade"
elif score < 0.8:
return "USA 3rd Grade"
elif score < 1.3:
return "USA 4th Grade"
elif score < 1.8:
return "USA 5th Grade"
elif score < 2.4:
return "USA 6th Grade"
elif score < 3.0:
return "USA 7th Grade"
elif score < 3.7:
return "USA 8th Grade"
elif score < 4.5:
return "USA 9th Grade"
elif score < 5.3:
return "USA 10th Grade"
elif score < 6.2:
return "USA 11th Grade"
elif score < 7.2:
return "USA 12th Grade"
else:
return "USA College Level"
#DISTRIBUTIONS_______________________________________________________________________
def get_word_length_distribution(tokens):
"""
Compute word length distribution
Arguments:
tokens: list
Returns:
{token_length: rel_freq}
"""
lengths = [len(t) for t in tokens]
dist = dict(Counter(lengths))
dist = {k:v/len(tokens) for k,v in sorted(dist.items(), key=lambda x: x[0])}
dist = {int(k): [v] for k,v in dist.items()}
return dist
def get_dependency_distribution(dependencies):
"""
Compute dependency distribution
Arguments:
dependencies: list
Returns:
{dependency: rel_freq}
"""
dist = dict(Counter(dependencies))
dist = {k:v/len(dependencies) for k,v in dist.items()}
dist = {k: [v] for k,v in dist.items()}
return dist
# def get_grapheme_distribution(tokens):
# """
# Compute grapheme distribution
# Arguments:
# tokens: lst
# Returns:
# {grapheme: rel_freq}
# """
# graphemes = ''.join(tokens)
# n_total = len(graphemes)
# dist = dict(Counter(graphemes))
# for k,v in dist.items():
# dist[k] = v/n_total
# dist = dict(sorted(dist.items(), key=operator.itemgetter(1),reverse=True))
# return dist
# def get_word_internal_grapheme_profile(tokens):
# """
# Compute word-interal grapheme distribution
# Arguments:
# tokens: lst
# Returns:
# {grapheme: % of wors that contain grapheme}
# """
# graphemes = set(''.join(tokens))
# n_tokens = len(tokens)
# profile = {}
# for g in graphemes:
# for t in tokens:
# if g in t:
# if g in profile.keys():
# profile[g+'_word_internal'] += 1
# else:
# profile[g+'_word_internal'] = 1
# profile = {k:v/n_tokens for k,v in profile.items()}
# profile = dict(sorted(profile.items(), key=operator.itemgetter(1),reverse=True))
# return profile
def get_function_word_distribution(doc):
"""
Compute function word distribution
Arguments:
doc: Stanza doc object
Returns:
{function word: rel_freq}
"""
allowed_pos = {'ADP', 'AUX', 'CCONJ', 'DET', 'PART', 'PRON', 'SCONJ'}
function_words = [w.text.lower() for s in doc.sents for w in s if w.pos_ in allowed_pos]
n_function_words = len(function_words)
dist = dict(Counter(function_words))
dist = {k:v/n_function_words for k,v in dist.items()}
dist = {k: [v] for k,v in dist.items()}
dist = dict(sorted(dist.items(), key=operator.itemgetter(1),reverse=True))
return dist
# def get_grapheme_positional_freq(tokens):
# """
# Compute positional frequency of graphemes
# Arguments:
# tokens: lst
# Returns:
# {pos_idx_in_token: {char: rel_freq}
# """
# n_tokens = len(tokens)
# lengths = [len(t) for t in tokens]
# longest = sorted(lengths)[-1]
# profile={}
# for i in range(1, longest+1):
# grapheme_freq = {}
# for t in tokens:
# if len(t) >= i:
# if t[i-1] in grapheme_freq.keys():
# grapheme_freq[t[i-1]] += 1
# else:
# grapheme_freq[t[i-1]] = 1
# grapheme_freq = {k:v/n_tokens for k,v in grapheme_freq.items()}
# grapheme_freq = {f'char_idx_{i}_{k}':v for k,v in grapheme_freq.items()}
# grapheme_freq = dict(sorted(grapheme_freq.items(), key=operator.itemgetter(1),reverse=True))
# profile['char_idx_'+str(i)] = grapheme_freq
# return profile
def get_punct_dist(text):
"""
Compute punctuation distribution
Arguments:
tokens: lst
Returns:
{punct: relative frequency by n characters}
"""
dist = {}
n_punct = 0
for p in punctuation:
n = text.count(p)
dist[p] = n
n_punct += n
if not n_punct:
return None
dist_by_char = {k: [v/n_punct] for k,v in dist.items()}
dist_by_char = dict(sorted(dist_by_char.items(), key=operator.itemgetter(1),reverse=True))
return dist_by_char
# def get_positional_word_profile(doc):
# """
# Compute positional word profile
# Arguments:
# doc: stanza doc object
# Returns:
# {token idx in sentence: {word: relative freq}}
# """
# tokens = [[w.text.lower() for w in s] for s in doc.sents]
# n_positions = max([len(s) for s in tokens])
# profile = {}
# for i in range(n_positions):
# k = i
# words = [s[k] for s in tokens if len(s)>k]
# n_sentences = len(words)
# v = dict(Counter(words))
# v = {k:v/n_sentences for k,v in v.items()}
# v = {f'token_idx_{str(i)}_{k}':v for k,v in v.items()}
# v = dict(sorted(v.items(), key=operator.itemgetter(1),reverse=True))
# profile['token_idx_'+str(k)] = v
# return profile
def get_ngram_profile(tokens):
"""
Compute ngram distribution
Arguments:
tokens: lst
ngram_range: (min, max)
Returns:
{ngram: freq}
"""
tokens = [[t.lower() for t in tokens]]
vec = CountVectorizer(analyzer=lambda x:x)
X = vec.fit_transform(tokens)
document_lengths = X.sum(axis=1)
X_normalized = X / document_lengths
profile = dict()
for v,k in zip(X_normalized.toarray().flatten(), vec.get_feature_names_out()):
profile[k] = [v]
return profile