-
Notifications
You must be signed in to change notification settings - Fork 7
/
single.py
438 lines (364 loc) · 15.7 KB
/
single.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
"""
Variant on Multi-Author Attribution System for Single-Author Attribution
Two evaluation methods:
- whole document vector
- sentence-level vectors and taking the plurality
"""
from collections import Counter
from types import SimpleNamespace
from sklearn.naive_bayes import GaussianNB
import argparse
import art
import glob
import importlib
import math
import nltk
import numpy as np
import os
import random
import shutil
import sys
import tabulate
import yaml
import markov
# (a, b, c) correspond tot hese values
TRAINING_DOCUMENTS_PER_AUTHOR = 10 # Train = n
TESTING_DOCUMENTS_PER_AUTHOR = 50
SENTENCES_PER_DOCUMENT = 20
VERBOSE = False
class Dataset():
"""
Two instantiations.
Whole Document Vector
Training [0]: (0, author, sentence_index)
Testing [1]: (1, document_index, 0) <- 0 because there will only ever be one sentence index
Sentence-Level Vectors
Training [0]: (0, author, sentence_index)
Testing [1]: (1, document_index, sentence_index)
"""
def __init__(self):
pass
def set_training(self, training):
self.training = training
def set_testing(self, testing):
self.testing = testing
def prepare(self):
"""Prepares dataset contents for passing into features and vectors."""
self.contents = dict()
self.vectors = dict()
for author in self.training:
for i, sentence in enumerate(self.training[author]):
self.contents[(0, author, i)] = sentence
self.vectors[(0, author, i)] = []
for i, composite in enumerate(self.testing):
for j, sentence in enumerate(composite):
self.contents[(1, i, j)] = sentence
self.vectors[(1, i, j)] = []
def add_vectors(self, vectors):
for identifier in vectors:
self.vectors[identifier].extend(vectors[identifier])
def training_labels(self):
"""Returns data, labels for fitting"""
data = []
labels = []
for author in self.training:
for i, _ in enumerate(self.training[author]):
data.append(self.vectors[(0, author, i)])
labels.append(author)
return data, labels
def predict(self, clf):
self.predictions = []
for i, composite in enumerate(self.testing):
vectors = []
for j, sentence in enumerate(composite):
vectors.append(self.vectors[(1, i, j)])
predictions = list(clf.predict(vectors))
self.predictions.append(predictions)
for j, sentence in enumerate(composite):
sentence.prediction = predictions[j]
class Runner():
def __init__(self, config):
self.log("Initializing runner...")
self.config = config
def log(self, message, override=False, clearline=False):
if override or VERBOSE:
if clearline:
sys.stdout.write("\033[F")
print(message)
def run(self):
self.clean()
self.preprocess()
self.load_sentences()
for i in range(10):
print(f"START INSTANCE {i}")
self.run_instance() # loop this line multiple times if needed
print(f"END INSTANCE {i}")
def run_instance(self):
self.log("Running instance...")
self.training = dict()
self.testing = dict()
self.generate() # split into training and testing sets
self.prepare_dataset()
self.train()
self.predict()
self.print_results()
def preprocess(self):
self.mkdir(os.path.join(self.config.src, self.config.sentence_dir))
if self.config.should_skip("preprocessing"):
self.log("Skipping preprocessing...")
return
self.log("Preprocessing...")
self.log("Generating sample documents...")
# Gather all sentences from the corpus.
sentences = {author: [] for author in self.config.authors}
for author_config in self.config.corpus:
author = author_config.author
src = glob.glob(author_config.src)
for filename in src:
contents = open(filename).read().replace("\n", " ")
sentences[author].extend(nltk.sent_tokenize(contents))
for author in self.config.authors:
author_identifier = self.config.authors[author]
digits = math.ceil(math.log10(len(sentences[author])))
for i, sentence in enumerate(sentences[author]):
dirname = f"{author_identifier}_{str(i).zfill(digits)}"
self.log(f"Preprocessing {dirname}...")
self.mkdir(os.path.join(self.config.src, self.config.sentence_dir, dirname))
for preprocessor in self.config.preprocessors:
filename = os.path.join(self.config.src, self.config.sentence_dir, dirname, preprocessor.name)
if os.path.exists(filename):
self.log(f" Skipping {preprocessor.name} for {dirname}, already exists...")
continue
self.log(f" Using preprocessor {preprocessor.name} for {dirname}...")
result = preprocessor.process(sentence)
f = open(filename, "w")
f.write(result)
f.close()
def load_sentences(self):
"""Loads forms of all sentences into memory, indexed using identifiers and storing a Sentence"""
self.log("Loading sentences...")
self.sentences = dict()
for i, author in enumerate(self.config.authors):
author_identifier = self.config.authors[author]
identifiers = glob.glob(os.path.join(self.config.src, self.config.sentence_dir, f"{author_identifier}_*"))
for j, identifier in enumerate(identifiers):
self.log(f" Loading sentence {j + 1} of {len(identifiers)} for author {i+1} of {len(self.config.authors)}...", clearline=i > 0 or j > 0)
sentence = Sentence(author, identifier)
self.sentences[identifier] = sentence
def generate(self):
def generate_document_from_sentences(sentences, n):
sampled = []
for i in range(n):
k = random.randint(0, len(sentences) - 1)
sampled.append(sentences[k])
del sentences[k]
return sampled
self.training_documents = {}
self.testing_documents = []
for author in self.config.authors:
self.training_documents[author] = []
sentences = { sentence: self.sentences[sentence] for sentence in self.sentences if self.sentences[sentence].author == author }
identifiers = list(sentences.keys())
for i in range(TRAINING_DOCUMENTS_PER_AUTHOR):
sampled = generate_document_from_sentences(identifiers, SENTENCES_PER_DOCUMENT)
sentences = [self.sentences[identifier] for identifier in sampled]
document = Document(author, f"Training Document {i}", sentences)
self.training_documents[author].append(document)
for i in range(TESTING_DOCUMENTS_PER_AUTHOR):
sampled = generate_document_from_sentences(identifiers, SENTENCES_PER_DOCUMENT)
sentences = [self.sentences[identifier] for identifier in sampled]
document = Document(author, f"Testing Document {i}", sentences)
self.testing_documents.append(document)
self.log(f"Done generating documents.")
def prepare_dataset(self):
self.log("Preparing datasets...")
# Loading sentences into two datasets, one for the whole dataset and one for sentences.
self.whole_dataset = Dataset()
self.whole_dataset.set_training(self.training_documents)
self.whole_dataset.set_testing([[document] for document in self.testing_documents])
self.whole_dataset.prepare()
self.sentence_dataset = Dataset()
training = {}
for author in self.training_documents:
training[author] = []
for document in self.training_documents[author]:
training[author].extend(document.sentences)
testing = []
for document in self.testing_documents:
testing.append(document.sentences)
self.sentence_dataset.set_training(training)
self.sentence_dataset.set_testing(testing)
self.sentence_dataset.prepare()
print("Done preparing dataset.")
def train(self):
self.log("Computing vectors...")
for feature in self.config.features:
self.log(f"Training feature {feature.name} ({feature.config})...")
vectors = feature.train(self.whole_dataset)
vectors = feature.train(self.sentence_dataset)
self.log("Fitting...")
data, labels = self.whole_dataset.training_labels()
self.whole_clf = GaussianNB()
self.whole_clf.fit(data, labels)
data, labels = self.sentence_dataset.training_labels()
self.sentence_clf = GaussianNB()
self.sentence_clf.fit(data, labels)
def predict(self):
# Do the actual sentence prediction.
self.log("Predicting...")
self.whole_dataset.predict(self.whole_clf)
self.sentence_dataset.predict(self.sentence_clf)
def print_results(self):
(overall_accurate, overall_inaccurate) = (0, 0)
(sentence_accurate, sentence_inaccurate) = (0, 0)
data = []
for i in range(len(self.whole_dataset.testing)):
author = self.whole_dataset.testing[i][0].author
whole_prediction = self.whole_dataset.testing[i][0].prediction
sentence_counts = Counter([sentence.prediction for sentence in self.sentence_dataset.testing[i]])
sentence_prediction = sentence_counts.most_common(1)[0][0]
data.append([i, author, whole_prediction, sentence_prediction, sentence_counts])
if author == whole_prediction:
overall_accurate += 1
else:
overall_inaccurate += 1
if author == sentence_prediction:
sentence_accurate += 1
else:
sentence_inaccurate += 1
print(tabulate.tabulate(data, headers=["Identifier", "Author", "Whole Prediction", "Sentence Prediction", "Counts"], tablefmt="psql"))
overall_accuracy = "{:.2f}%".format((overall_accurate / (overall_accurate + overall_inaccurate)) * 100)
sentence_accuracy = "{:.2f}%".format((sentence_accurate / (sentence_accurate + sentence_inaccurate)) * 100)
print(f"Overall Accuracy: {overall_accurate} of {overall_accurate + overall_inaccurate} ({overall_accuracy})")
print(f"Sentence Accuracy: {sentence_accurate} of {sentence_accurate + sentence_inaccurate} ({sentence_accuracy})")
def mkdir(self, dirname):
try:
os.makedirs(dirname)
except FileExistsError:
pass
def clean(self):
for dirname in self.config.clean:
self.log(f"Cleaning up directory {dirname}...")
shutil.rmtree(dirname, ignore_errors=True)
class Config():
def __init__(self, contents):
self.config = contents
self.src = self.config["configuration"]["src"]
self.sentence_dir = self.config["configuration"]["sentence_dir"]
self.training_threshold = self.config["configuration"]["train"]
self.accuracy = self.config["configuration"]["accuracy"]
self.corpus = self.get_corpus()
self.authors = self.get_authors()
self.generate = self.get_generate_prob()
self.features = self.get_features()
self.preprocessors = self.get_preprocessors()
self.clean = self.get_clean()
def get_corpus(self):
data = []
for author_config in self.config["corpus"]:
config = SimpleNamespace()
config.author = author_config["author"]
config.src = os.path.join(self.src, author_config["src"])
data.append(config)
return data
def get_authors(self):
data = {}
for author in self.config["authors"]:
data[author] = self.config["authors"][author]
return data
def get_generate_prob(self):
data = SimpleNamespace()
for feature in ["stay", "terminate", "threshold", "n"]:
setattr(data, feature, self.config["generate"][feature])
data.next = 1 - data.stay - data.terminate
return data
def get_features(self):
data = []
for feature in self.config["features"]:
data.append(Feature(feature))
return data
def get_preprocessors(self):
preprocessors = []
for preprocessor in self.config["preprocessors"]:
preprocessors.append(Preprocessor(preprocessor))
return preprocessors
def get_clean(self):
return [os.path.join(self.src, dirname) for dirname in self.config["configuration"]["clean"]]
def should_skip(self, step):
return step in self.config["configuration"].get("skip", [])
class Feature():
def __init__(self, feature):
try:
self.name = feature["name"]
self.module = importlib.import_module(f"features.{self.name}")
self.config = feature
except ModuleNotFoundError:
raise Exception(f"Could not find feature {self.name}.")
def train(self, dataset):
contents = dataset.contents
vectors = self.module.train(self.config, contents)
dataset.add_vectors(vectors)
class Preprocessor():
def __init__(self, name):
try:
self.name = name
self.module = importlib.import_module(f"preprocessors.{name}")
except ModuleNotFoundError:
raise Exception(f"Could not find preprocessor {name}.")
def process(self, contents):
return self.module.preprocess(contents)
class Sentence():
def __init__(self, author, dirname):
self.author = author
self.identifier = dirname
attributes = os.listdir(dirname)
self.attributes = { attribute: open(os.path.join(dirname, attribute)).read().rstrip("\n")
for attribute in attributes }
def __getattr__(self, attr):
if attr in self.attributes:
return self.attributes[attr]
else:
print(attr)
print(self.attributes)
print(attr in self.attributes)
raise AttributeError
def get(self, attr):
return self.__getattr__(attr)
class Document(Sentence):
"""
A document is just a list of sentences.
"""
def __init__(self, author, identifier, sentences):
self.author = author
self.identifier = identifier
self.sentences = sentences
# Combine sentences attributes.
self.attributes = {}
for sentence in self.sentences:
for attr in sentence.attributes:
if attr not in self.attributes:
self.attributes[attr] = sentence.attributes[attr]
else:
self.attributes[attr] += "\n"
self.attributes[attr] += sentence.attributes[attr]
def main():
config = parse_config()
runner = Runner(config)
runner.run()
def parse_config():
global VERBOSE
parser = argparse.ArgumentParser(
description="Run a multi-authorship attribution classifier on anonymous documents."
)
parser.add_argument("config", type=str)
parser.add_argument("-v", "--verbose", action="store_true")
args = parser.parse_args()
if args.verbose:
VERBOSE = True
contents = open(args.config).read()
data = yaml.load(contents)
return Config(data)
if __name__ == "__main__":
print(art.text2art("Authorship"))
print(art.text2art("Attribution"))
main()