-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathauthorship_attribution_n_authors_sentence_classification.py
182 lines (156 loc) · 7.53 KB
/
authorship_attribution_n_authors_sentence_classification.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
#!/usr/bin/env python
# -*- coding: utf-8 -*-
#
# File : core.classifiers.RCNLPTextClassifier.py
# Description : Echo State Network for text classification.
# Auteur : Nils Schaetti <nils.schaetti@unine.ch>
# Date : 01.02.2017 17:59:05
# Lieu : Nyon, Suisse
#
# This file is part of the Reservoir Computing NLP Project.
# The Reservoir Computing Memory Project is a set of free software:
# you can redistribute it and/or modify
# it under the terms of the GNU General Public License as published by
# the Free Software Foundation, either version 3 of the License, or
# (at your option) any later version.
#
# Foobar is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
# GNU General Public License for more details.
# You should have received a copy of the GNU General Public License
# along with Foobar. If not, see <http://www.gnu.org/licenses/>.
#
import io
import os
import argparse
import matplotlib.pyplot as plt
import pickle
import numpy as np
import Oger
import spacy
from core.converters.RCNLPPosConverter import RCNLPPosConverter
from core.converters.RCNLPTagConverter import RCNLPTagConverter
from core.converters.RCNLPWordVectorConverter import RCNLPWordVectorConverter
from core.converters.RCNLPFuncWordConverter import RCNLPFuncWordConverter
from core.classifiers.RCNLPEchoWordClassifier import RCNLPEchoWordClassifier
from core.tools.RCNLPLogging import RCNLPLogging
#########################################################################
#
# Experience settings
#
#########################################################################
# Exp. info
ex_name = "Authorship Attribution Experience"
ex_instance = "Author Attribution n authors sentence classification"
# Reservoir Properties
rc_leak_rate = 0.1 # Leak rate
rc_input_scaling = 0.25 # Input scaling
rc_size = 100 # Reservoir size
rc_spectral_radius = 0.99 # Spectral radius
rc_w_sparsity = 0.1
rc_input_sparsity = 0.1
####################################################
# Main function
####################################################
if __name__ == "__main__":
# Argument parser
parser = argparse.ArgumentParser(description="RCNLP - Authorship attribution with Echo State Network")
# Argument
parser.add_argument("--dataset", type=str, help="Dataset's directory.")
parser.add_argument("--author1", type=int, help="Author 1' ID.")
parser.add_argument("--author2", type=int, help="Author 2's ID.")
parser.add_argument("--samples", type=int, help="Number of samples to use to assess accuracy.", default=20)
parser.add_argument("--lang", type=str, help="Language (ar, en, es, pt)", default='en')
parser.add_argument("--converter", type=str, help="The text converter to use (fw, pos, tag, wv).", default='pos')
parser.add_argument("--pca-model", type=str, help="PCA model to load", default=None)
parser.add_argument("--in-components", type=int, help="Number of principal component to reduce inputs to.",
default=-1)
parser.add_argument("--k", type=int, help="n-Fold Cross Validation.", default=10)
args = parser.parse_args()
# Logging
logging = RCNLPLogging(exp_name=ex_name, exp_inst=ex_instance,
exp_value=RCNLPLogging.generate_experience_name(locals()))
logging.save_globals()
logging.save_variables(locals())
# PCA model
pca_model = None
if args.pca_model != "":
pca_model = pickle.load(open(args.pca_model, 'r'))
# end if
# >> 1. Choose a text to symbol converter.
if args.converter == "pos":
converter = RCNLPPosConverter(resize=args.in_components, pca_model=pca_model)
elif args.converter == "tag":
converter = RCNLPTagConverter(resize=args.in_components, pca_model=pca_model)
elif args.converter == "fw":
converter = RCNLPFuncWordConverter(resize=args.in_components, pca_model=pca_model)
else:
converter = RCNLPWordVectorConverter(resize=args.in_components, pca_model=pca_model)
# end if
# >> 2. Prepare training and test set indexes.
n_fold_samples = int(100 / args.k)
indexes = np.arange(0, 100, 1)
indexes.shape = (args.k, n_fold_samples)
# >> 3. Array for results
average_success_rate = np.array([])
# >> 4. n-Fold cross validation
for k in range(0, args.k):
print("%d-Fold" % k)
# >> 5. Prepare training and test set.
test_set_indexes = indexes[k]
training_set_indexes = indexes
training_set_indexes = np.delete(training_set_indexes, k, axis=0)
training_set_indexes.shape = (100 - n_fold_samples)
# >> 6. Create Echo Word Classifier
classifier = RCNLPEchoWordClassifier(size=rc_size, input_scaling=rc_input_scaling, leak_rate=rc_leak_rate,
input_sparsity=rc_input_sparsity, converter=converter, n_classes=2,
spectral_radius=rc_spectral_radius, w_sparsity=rc_w_sparsity)
# >> 7. Add examples
print(training_set_indexes)
for author_index, author_id in enumerate((args.author1, args.author2)):
author_path = os.path.join(args.dataset, "total", str(author_id))
print("Adding %d examples for author from %s" % (training_set_indexes.shape[0], author_path))
for file_index in training_set_indexes:
classifier.add_example(os.path.join(author_path, str(file_index) + ".txt"), author_index)
# end for
# end for
# >> 8. Train model
print("Training model with text files from %s" % os.path.join(args.dataset, "train"))
classifier.train()
# >> 9. Test sentence classification success rate
print("Testing sentence classification with text files from %s..." % os.path.join(args.dataset, "total"))
print(test_set_indexes)
success = 0.0
count = 0.0
for author_index, author_id in enumerate((args.author1, args.author2)):
author_path = os.path.join(args.dataset, "total", str(author_id))
print("Testing model performances with %d text files for author from %s..." % (test_set_indexes.shape[0],
author_path))
# Iterate over files
for file_index in test_set_indexes:
# Load text file
nlp = spacy.load(args.lang)
doc = nlp(io.open(os.path.join(author_path, str(file_index) + ".txt"), 'r').read())
# Iterate over sentences
for sentence in doc.sents:
sentence_pred = classifier.pred_text(sentence.text)
if sentence_pred == author_index:
success += 1.0
# end if
count += 1.0
# end for
# end for
# >> 10. Log success
logging.save_results("Number of sentences in test set ", count, display=True)
logging.save_results("Number of successes ", success, display=True)
logging.save_results("Success rate ", (success / count) * 100.0, display=True)
# >> 11. Save results
average_success_rate = np.append(average_success_rate, [(success / count) * 100.0])
# Delete variables
del classifier
# end for
# Log results
logging.save_results("Average success rate ", np.average(average_success_rate), display=True)
logging.save_results("Success rate std ", np.std(average_success_rate), display=True)
# end if