-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathauthorship_attribution_two_authors_one_hot_inputs.py
179 lines (152 loc) · 6.93 KB
/
authorship_attribution_two_authors_one_hot_inputs.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
#!/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 pickle
import numpy as np
import spacy
from scipy import stats
from core.converters.PosConverter import PosConverter
from core.converters.TagConverter import TagConverter
from core.converters.WVConverter import WVConverter
from core.converters.FuncWordConverter import FuncWordConverter
from core.converters.OneHotConverter import OneHotConverter
from core.classifiers.EchoWordClassifier import EchoWordClassifier
from core.tools.RCNLPLogging import RCNLPLogging
from core.tools.Metrics import Metrics
import logging
from core.embeddings.Word2Vec import Word2Vec
#########################################################################
# Experience settings
#########################################################################
# Exp. info
ex_name = "Authorship Attribution"
ex_instance = "Two Authors One-hot representations"
# Reservoir Properties
rc_leak_rate = 0.5 # Leak rate
rc_input_scaling = 1.0 # Input scaling
rc_size = 4000 # Reservoir size
rc_spectral_radius = 0.99 # Spectral radius
rc_w_sparsity = 0.1
rc_input_sparsity = 0.005
####################################################
# Functions
####################################################
####################################################
# Main function
####################################################
if __name__ == "__main__":
# Argument parser
parser = argparse.ArgumentParser(
description="RCNLP - Compare the Echo Text Classifier to other models with two authors")
# Argument
parser.add_argument("--dataset", type=str, help="Dataset's directory")
parser.add_argument("--author1", type=str, help="First author", default="1")
parser.add_argument("--author2", type=str, help="Second author", default="2")
parser.add_argument("--lang", type=str, help="Language (en_core_web_md, ar, en, es, pt)", default='en_core_web_md')
parser.add_argument("--sentence", action='store_true', help="Test sentence classification rate?", default=False)
parser.add_argument("--k", type=int, help="n-Fold Cross Validation", default=10)
parser.add_argument("--verbose", action='store_true', help="Verbose mode", default=False)
parser.add_argument("--debug", action='store_true', help="Debug mode", default=False)
parser.add_argument("--voc-size", type=int, help="Vocabulary size", default=5000, required=True)
parser.add_argument("--log-level", type=int, help="Log level", default=20)
parser.add_argument("--sparse", action='store_true', help="Sparse matrix?", default=False)
parser.add_argument("--multi", action='store_true', help="Probabilities computer by multiplication not average", default=False)
args = parser.parse_args()
# Init logging
logging.basicConfig(level=args.log_level)
logger = logging.getLogger(name="RCNLP")
# Word2Vec
word2vec = Word2Vec(dim=args.voc_size, mapper='one-hot')
# Choose a text to symbol converter
converter = OneHotConverter(lang=args.lang, voc_size=args.voc_size, word2vec=word2vec)
# 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)
# Aggregation
if args.multi:
aggregation = 'multiplication'
else:
aggregation = 'average'
# end if
# Create Echo Word Classifier
classifier = EchoWordClassifier(classes=[0, 1], size=rc_size, input_scaling=rc_input_scaling,
leak_rate=rc_leak_rate,
input_sparsity=rc_input_sparsity, converter=converter,
spectral_radius=rc_spectral_radius, w_sparsity=rc_w_sparsity,
use_sparse_matrix=args.sparse, aggregation=aggregation)
# Success rates
success_rates = np.zeros(args.k)
# k-Fold cross validation
for k in range(0, args.k):
# 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)
# Add examples
for author_index, author_id in enumerate((args.author1, args.author2)):
author_path = os.path.join(args.dataset, "total", author_id)
for file_index in training_set_indexes:
file_path = os.path.join(author_path, str(file_index) + ".txt")
classifier.train(io.open(file_path, 'r').read(), author_index)
# end for
# end for
# Finalize model training
classifier.finalize(verbose=args.verbose)
# Display embeddings
print(classifier.get_embeddings())
# Init test epoch
test_set = list()
# Get text
for author_index, author_id in enumerate((args.author1, args.author2)):
author_path = os.path.join(args.dataset, "total", str(author_id))
for file_index in test_set_indexes:
file_path = os.path.join(author_path, str(file_index) + ".txt")
# Document success rate
if not args.sentence:
test_set.append((io.open(file_path, 'r').read(), author_index))
else:
# Sentence success rate
nlp = spacy.load(args.lang)
doc = nlp(io.open(file_path, 'r').read())
for sentence in doc.sents:
test_set.append((sentence, author_index))
# end for
# end if
# end for
# end for
# Success rate
success_rate = Metrics.success_rate(classifier, test_set, verbose=args.verbose, debug=args.debug)
logger.info(u"\t{} - Success rate : {}".format(k, success_rate))
# Save result
success_rates[k] = success_rate
# Reset
classifier.reset()
# end for
# Over all success rate
logger.info(u"All - Success rate : {}".format(np.average(success_rates)))
# end if