-
Notifications
You must be signed in to change notification settings - Fork 0
/
localmaxs.py
518 lines (388 loc) · 14.3 KB
/
localmaxs.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
# -*- coding: utf-8 -*-
"""
Created on Tue May 3 18:54:37 2022
@author: eugen
"""
import re
import os
import numpy as np
import math
import pickle
import random
import matplotlib.pyplot as plt
from nltk import FreqDist
from nltk.util import ngrams
from zipfile import ZipFile
from sklearn.feature_extraction.text import CountVectorizer
corpus = []
archive = ZipFile('corpus2mw.zip', 'r')
fileList = archive.namelist()
for file in fileList:
corpus.append((archive.read(file)).decode('UTF-8'))
regexp = re.compile('[\w \-]')
def dice(freq, pref_freqs, suff_freqs):
pref_freqs = list(pref_freqs)
suff_freqs = list(suff_freqs)
return 2 * freq / (sum(pref_freqs) / len(pref_freqs) + sum(suff_freqs) / len(suff_freqs))
def scp(freq, pref_freqs, suff_freqs):
multiplied_freqs = [pref_freq * suff_freq for pref_freq, suff_freq in zip(pref_freqs, suff_freqs)]
return freq ** 2 * len(multiplied_freqs) / sum(multiplied_freqs)
def processText(corpus):
corp = []
for text in corpus:
listT = list(text)
i = 0
for c in listT:
if (not regexp.search(c) and not listT[i-1]==' ') or not regexp.search(listT[i-1]) and regexp.search(c):
listT.insert(i, ' ')
i +=1
corp.append(''.join(listT))
return corp
readPickle = True
if readPickle:
with open('corpusList', 'rb') as fp:
corpus = pickle.load(fp)
else:
corpus = processText(corpus)
with open('corpusList', 'wb') as fp:
pickle.dump(corpus, fp)
def compute_freq_doc(text, minG, maxG):
freq_dist = FreqDist()
if len(text) > 1:
tokens = text.strip().split()
for i in range(minG, maxG+1):
grams = ngrams(tokens, i)
freq_dist.update(grams)
return dict(freq_dist)
def compute_freq_corpus(minG, maxG):
freq_dist = FreqDist()
for text in corpus:
if len(text) > 1:
tokens = text.strip().split()
for i in range(minG, maxG+1):
grams = ngrams(tokens, i)
freq_dist.update(grams)
return dict(freq_dist)
freq_dict = compute_freq_corpus(1, 8)
#transform tuple keys to string and filter all ngrams value > 1
freq_dict = {' '.join(key):val for key, val in freq_dict.items() if val > 1}
filtered_dict_sorted= sorted(freq_dict.items(), key=lambda x: len(x[0].split()), reverse=True)
#unigrams
vectorizer = CountVectorizer(token_pattern=r'\w+')
vec_fit = vectorizer.fit_transform(corpus)
single_word_list = vectorizer.get_feature_names_out()
single_count_list = np.asarray(vec_fit.sum(axis=0))[0]
single_freq_dict = dict(zip(single_word_list,single_count_list))
single_freq_dict = {key:val for key, val in single_freq_dict.items() if val > 1}
single_freq_dict = {k: v for k, v in sorted(single_freq_dict.items(), key=lambda item: item[1])}
values = np.fromiter(single_freq_dict.values(), dtype=float)
stop_words_list = np.stack((np.arange(0, len(single_freq_dict)), values), axis = -1)
list_of_counts = list(single_freq_dict.items())
from kneebow.rotor import Rotor
rotor = Rotor()
rotor.fit_rotate(stop_words_list)
elbow_idx = rotor.get_elbow_index()
print(elbow_idx)
#rotor.plot_elbow()
from kneed import KneeLocator
kn = KneeLocator(stop_words_list[:,0] ,stop_words_list[:,1], curve='convex', direction='increasing')
print(int(kn.knee))
stop = 0
deltaX = 200
for idx, i, j in zip(range(0, len(values)), values, values[deltaX:]):
if((j-i)>stop): stop = idx
print(stop)
fig = plt.figure()
ax = plt.gca()
ax.scatter(stop_words_list[:,0] ,stop_words_list[:,1] , s=1,c='blue', marker='.')
ax.set_yscale('log')
#ax.set_xscale('log')
fig.set_size_inches(10, 7)
plt.axvline(x=stop, color='r', linestyle='--')
expressions_count={}
for key, val in filtered_dict_sorted:
expressions_count[key] = val
stop_words_unigrams = list_of_counts[stop:]
stop_words=[]
for key, val in stop_words_unigrams:
stop_words.append(key)
'''
poss_re={}
for key, val in filtered_dict_sorted:
words= key.split()
n= len(words)
if len(words) > 1:
ownpref=''
ownsuf=''
for i in range(0,n):
if i==0:
ownpref += key.split(' ')[i]
elif i==n-1:
ownsuf += key.split(' ')[i]
else:
ownpref += ' ' + key.split(' ')[i]
ownsuf += key.split(' ')[i] + ' '
xpref= expressions_count[ownpref]
xsuf= expressions_count[ownsuf]
scpg= val**2 /(xpref * xsuf)
diceg = 2 * val / (xpref + xsuf)
poss_re[key] = {'n':n, 'freq': val, 'scpg': scpg, 'diceg': diceg,'xpref': ownpref, 'xsuf': ownsuf}
with open('PossREList', 'wb') as fp:
pickle.dump(poss_re, fp)
'''
with open('PossREList', 'rb') as fp:
poss_re = pickle.load(fp)
#x = best glue from n-1 words
'''
points=['.',',','?','(',')','!','@','&','^','~','|','>','<', '%', '$', '[', ']', '{', '}', ':', ';', '-','a','s', '_', '+', '=', '*', '\\', '\'', '\"', '`', '#', '0', '1', '2', '3', '4', '5', '6', '7', '8', '9', '/', '-', "'", 't']
relevant_expressions={}
res=[]
for key, val in poss_re.items():
if val['n'] > 2:
#print(val['n'], 'ngram')
words = key.split()
if words[0].lower() in stop_words:
continue
if words[-1].lower() in stop_words:
continue
if val['freq'] <= 2:
continue
grams = key.split()
puntcount = 0
for g in grams:
if g in points:
puntcount+=1
if puntcount!=0:
continue
#best x
try:
xpref = poss_re[val['xpref']]['scpg']
xsuf = poss_re[val['xsuf']]['scpg']
bestxscp = max(xpref, xsuf)
xpref = poss_re[val['xpref']]['diceg']
xsuf = poss_re[val['xsuf']]['diceg']
bestxdice = max(xpref, xsuf)
except:
continue
#best y
bestyscp=1
for keyy, valy in poss_re.items():
if key in keyy and valy['n'] == val['n']+1:
if valy['scpg']<bestyscp:
bestyscp=valy['scpg']
bestydice=1
for keyy, valy in poss_re.items():
if key in keyy and valy['n'] == val['n']+1:
if valy['diceg']<bestydice:
bestydice=valy['diceg']
#see if is relevant
if (val['scpg']-0.6)>(bestxscp+bestyscp)/2 and (val['diceg']-0.6)>(bestxdice+bestydice)/2:
print(key)
relevant_expressions[key]={'n':val['n'], 'freq': val['freq'], 'scpg': val['scpg'], 'xscp':bestxscp, 'yscp':bestyscp, 'diceg': val['diceg'], 'xdice':bestxdice, 'ydice':bestydice}
res.append(key)
else:
words = key.split()
if words[0].lower() in stop_words:
continue
if words[-1].lower() in stop_words:
continue
if val['freq'] <= 2:
continue
grams = key.split()
puntcount = 0
for g in grams:
if g in points:
puntcount+=1
if puntcount!=0:
continue
#print(val['n'], 'bigram')
for keyy, valy in poss_re.items():
if key in keyy and valy['n'] == val['n']+1:
if valy['scpg']<bestyscp:
bestyscp=valy['scpg']
bestydice=1
for keyy, valy in poss_re.items():
if key in keyy and valy['n'] == val['n']+1:
if valy['diceg']<bestydice:
bestydice=valy['diceg']
if (val['scpg']-0.6)>bestyscp and (val['diceg']-0.6)>bestydice:
#print(key)
relevant_expressions[key]={'n':val['n'], 'freq': val['freq'], 'scpg': val['scpg'], 'yscp':bestyscp, 'diceg': val['diceg'], 'ydice':bestydice}
res.append(key)
with open('REList', 'wb') as fp:
pickle.dump(relevant_expressions, fp)
'''
with open('REList', 'rb') as fp:
relevant_expressions = pickle.load(fp)
print('Precision')
print(100*len(relevant_expressions)/len(poss_re))
recallList = []
for i in range(0,200):
recallList.append(random.choice(list(poss_re.keys())))
countR = 0
for k in recallList:
if k in relevant_expressions.keys():
countR += 1
print('Recall')
print(100*countR/200)
print('F score')
print(100*2*((len(relevant_expressions)/len(poss_re))*(countR/200))/((len(relevant_expressions)/len(poss_re))+(countR/200)))
def count_RE_in_doc(RE):
count = 0
for text in corpus:
if RE in text:
count += 1
return count
def freq(RE,doc):
freq_dict = compute_freq_doc(doc, len(RE.split()), len(RE.split()))
if len(RE.split()) > 1:
freq_dict = {' '.join(key):val for key, val in freq_dict.items()}
else:
freq_dict = {''.join(key):val for key, val in freq_dict.items()}
return freq_dict[RE]
def tf_idf(RE, doc_idx):
doc = corpus[doc_idx]
freq_RE = freq(RE,doc)
return (freq_RE/len(doc.strip().split()))*math.log(len(corpus)/count_RE_in_doc(RE))
def findWholeWord(w):
return re.compile(r'\b({0})\b'.format(re.escape(w))).search
def calc_prob(word):
sum_p = 0
for doc in corpus:
if findWholeWord(word)(doc):
sum_p += freq(word, doc)/len(doc.strip().split())
return (1/len(corpus))*sum_p
def calc_cov(A,B):
probA = calc_prob(A)
probB = calc_prob(B)
sum_p = 0
for doc in corpus:
if findWholeWord(A)(doc) and findWholeWord(B)(doc):
sum_p += (freq(A, doc)/len(doc.strip().split())-probA)*(freq(B, doc)/len(doc.strip().split())-probB)
return (1/(len(corpus)-1))*sum_p
def correlation(A,B):
return calc_cov(A, B)/(math.sqrt(calc_cov(A, A))*math.sqrt(calc_cov(B, B)))
def get_distances(A,B,doc):
closest = 0
farthest = 0
listA = A.split()
listB = B.split()
listDoc = doc.strip().split()
idx_pos_A_1 = [ i for i in range(len(listDoc)) if listDoc[i] == listA[0] ]
idx_pos_A_2 = [ i for i in range(len(listDoc)) if listDoc[i] == listA[-1] ]
idx_pos_B_1 = [ i for i in range(len(listDoc)) if listDoc[i] == listB[0] ]
idx_pos_B_2 = [ i for i in range(len(listDoc)) if listDoc[i] == listB[-1] ]
idx_pos_A_1_copy = idx_pos_A_1
idx_pos_A_2_copy = idx_pos_A_2
idx_pos_B_1_copy = idx_pos_B_1
idx_pos_B_2_copy = idx_pos_B_2
for pos, idx in enumerate(idx_pos_A_1_copy):
for i, elem in enumerate(listA):
if listDoc[idx+i] != elem:
idx_pos_A_1.pop(pos)
break
for pos, idx in enumerate(idx_pos_A_2_copy):
for i, elem in enumerate(reversed(listA)):
if listDoc[idx-i] != elem:
idx_pos_A_2.pop(pos)
break
for pos, idx in enumerate(idx_pos_B_1_copy):
for i, elem in enumerate(listB):
if listDoc[idx+i] != elem:
idx_pos_B_1.pop(pos)
break
for pos, idx in enumerate(idx_pos_B_2_copy):
for i, elem in enumerate(reversed(listB)):
if listDoc[idx-i] != elem:
idx_pos_B_2.pop(pos)
break
listF = []
for a in idx_pos_A_1:
for b in idx_pos_B_2:
listF.append(a-b)
for a in idx_pos_A_2:
for b in idx_pos_B_1:
listF.append(b-a)
listF = [ i for i in listF if i > 0 ]
if len(listF) == 0: return 1
return min(listF)/max(listF)
def IP(A,B):
count = 0
sum_dist = 0
for i, doc in enumerate(corpus):
if findWholeWord(A)(doc) and findWholeWord(B)(doc):
count += 1
sum_dist += get_distances(A, B, doc)
return 1-(1/count)*sum_dist
def sem_prox(A,B):
return correlation(A, B)*math.sqrt(IP(A,B))
def occur_in_any_doc(A,B):
for doc in corpus:
if findWholeWord(A)(doc) and findWholeWord(B)(doc):
return True
return False
def calc_comp_medio(RE):
count = 0
listRE = RE.split()
for w in listRE:
count += len(w)
return count/len(listRE)
def score_explicit(doc_idx):
doc = corpus[doc_idx]
#unigrams
uni_dict = compute_freq_doc(doc, 1, 1)
uni_dict = {''.join(key):val for key, val in uni_dict.items()}
for k in uni_dict.keys():
uni_dict[k] = tf_idf(k,doc_idx) * calc_comp_medio(k)
top5_uni = dict(sorted(uni_dict.items(), key=lambda x: x[1], reverse=True)[:5])
#relevant expressions
re_dict = {}
for k in relevant_expressions.keys():
if k in doc:
re_dict[k] = relevant_expressions[k]['freq']
for k in re_dict.keys():
re_dict[k] = tf_idf(k,doc_idx) * calc_comp_medio(k)
top5_re = dict(sorted(re_dict.items(), key=lambda x: x[1], reverse=True)[:5])
return top5_uni , top5_re if len(re_dict) != 0 else 'No REs in the document'
def score_implicit(doc_idx):
doc = corpus[doc_idx]
#relevant expressions
re_dict_in = {}
re_dict_out = {}
for k in relevant_expressions.keys():
if k in doc:
re_dict_in[k] = relevant_expressions[k]['freq']
else:
re_dict_out[k] = relevant_expressions[k]['freq']
if len(re_dict_in) == 0: return 'No REs in the document'
for k in re_dict_in.keys():
re_dict_in[k] = tf_idf(k,doc_idx) * calc_comp_medio(k)
top10_re = dict(sorted(re_dict_in.items(), key=lambda x: x[1], reverse=True)[:10])
scores = {}
for k in re_dict_out.keys():
score = 0
for i, v in enumerate(top10_re):
if occur_in_any_doc(k, v):
score += sem_prox(k, v)/(i+1)
scores[k] = score
top10_scores = dict(sorted(scores.items(), key=lambda x: x[1], reverse=True)[:10])
return top10_scores
len_docs = {}
for i, doc in enumerate(corpus):
len_docs[i] = len(doc.strip().split())
#print(sorted(len_docs.items(), key=lambda x: x[1], reverse=True))
print("Doc 303")
print(score_explicit(303))
print(score_implicit(303))
print("Doc 1104")
print(score_explicit(1104))
print(score_implicit(1104))
print("Doc 1230")
print(score_explicit(1230))
print(score_implicit(1230))
print("Doc 1595")
print(score_explicit(1595))
print(score_implicit(1595))
print("Doc 2120")
print(score_explicit(2120))
print(score_implicit(2120))