-
Notifications
You must be signed in to change notification settings - Fork 1
/
mlp.py
218 lines (169 loc) · 7.13 KB
/
mlp.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
import sys
import json
from sklearn.externals import joblib
import numpy as np
import pickle
#Retrieval
import abc
from abc import abstractmethod
class Retrieval:
__metaclass__ = abc.ABCMeta
@classmethod
def __init__(self): #constructor for the abstract class
pass
@classmethod
def getLongSnippets(self, question):
longSnippets = question['contexts']['long_snippets']
fullLongSnippets = ' '.join(longSnippets)
return fullLongSnippets
@classmethod
def getShortSnippets(self, question):
shortSnippets = question['contexts']['short_snippets']
fullShortSnippets = ' '.join(shortSnippets)
return fullShortSnippets
#Featurization
import abc
from abc import abstractmethod
class Featurizer:
__metaclass__ = abc.ABCMeta
@classmethod
def __init__(self): #constructor for the abstract class
pass
#This is the abstract method that is implemented by the subclasses.
@abstractmethod
def getFeatureRepresentation(self, X_train, X_val):
pass
#TFIDF
#from Featurizer import Featurizer
from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.naive_bayes import MultinomialNB
#This is a subclass that extends the abstract class Featurizer.
class TfidfFeaturizer(Featurizer):
#The abstract method from the base class is implemeted here to return count features
def getFeatureRepresentation(self, X_train, X_val):
tfidf_vect = TfidfVectorizer(smooth_idf=True,min_df=20,max_df=2000)
X_train_counts = tfidf_vect.fit_transform(X_train)
X_val_counts = tfidf_vect.transform(X_val)
return X_train_counts, X_val_counts
#Count Featurizer
#from Featurizer import Featurizer
from sklearn.feature_extraction.text import CountVectorizer
from sklearn.naive_bayes import MultinomialNB
#This is a subclass that extends the abstract class Featurizer.
class CountFeaturizer(Featurizer):
#The abstract method from the base class is implemeted here to return count features
### got information about the CountVectorizer from Stack overflow answers - https://stackoverflow.com/questions/27697766/understanding-min-df-and-max-df-in-scikit-countvectorizer
def getFeatureRepresentation(self, X_train, X_val):
count_vect = CountVectorizer(min_df=20)
X_train_counts = count_vect.fit_transform(X_train)
X_val_counts = count_vect.transform(X_val)
return X_train_counts, X_val_counts
#Classifier
import abc
from abc import abstractmethod
class Classifier:
__metaclass__ = abc.ABCMeta
@classmethod
def __init__(self): #constructor for the abstract class
pass
#This is the abstract method that is implemented by the subclasses.
@abstractmethod
def buildClassifier(self, X_features, Y_train):
pass
#from Classifier import Classifier
from sklearn.naive_bayes import MultinomialNB
#This is a subclass that extends the abstract class Classifier.
class MultinomialNaiveBayes(Classifier):
#The abstract method from the base class is implemeted here to return multinomial naive bayes classifier
def buildClassifier(self, X_features, Y_train):
clf = MultinomialNB().fit(X_features, Y_train)
return clf
from sklearn.neural_network import MLPClassifier
class MLP(Classifier):
def buildClassifier(self,X_features,Y_features):
clf=MLPClassifier(solver='lbfgs', alpha=1e-5, hidden_layer_sizes=(50, 10), random_state=1).fit(X_features,Y_features)
return clf
from sklearn import svm
from sklearn.multiclass import OneVsRestClassifier
class SVM(Classifier):
def buildClassifier(self,X_features,Y_features):
clf=svm.LinearSVC(C=0.004).fit(X_features,Y_features)
return clf
#Evaluation
import abc
from abc import abstractmethod
import numpy as np
from sklearn.metrics import precision_recall_fscore_support
from sklearn.metrics import accuracy_score
class Evaluator:
__metaclass__ = abc.ABCMeta
@classmethod
def __init__(self): #constructor for the abstract class
pass
#This is a class method that gets accuracy of the model
@classmethod
def getAccuracy(self, Y_true, Y_pred):
accuracy = accuracy_score(Y_true, Y_pred)
return accuracy
#This is a class method that gets precision, recall and f-measure of the model
@classmethod
def getPRF(self, Y_true, Y_pred):
prf = precision_recall_fscore_support(Y_true, Y_pred, average='weighted')
precision = prf[0]
recall = prf[1]
f_measure = prf[2]
return precision, recall, f_measure
#PipelineQA
class Pipeline(object):
def __init__(self, trainFilePath, valFilePath, retrievalInstance, featurizerInstance, classifierInstance):
self.retrievalInstance = retrievalInstance
self.featurizerInstance = featurizerInstance
self.classifierInstance = classifierInstance
trainfile = open(trainFilePath, 'r')
self.trainData = json.load(trainfile)
trainfile.close()
valfile = open(valFilePath, 'r')
self.valData = json.load(valfile)
valfile.close()
self.question_answering()
def makeXY(self, dataQuestions):
X = []
Y = []
for question in dataQuestions:
long_snippets = self.retrievalInstance.getLongSnippets(question)
short_snippets = self.retrievalInstance.getShortSnippets(question)
X.append(short_snippets)
Y.append(question['answers'][0])
return X, Y
def question_answering(self):
dataset_type = self.trainData['origin']
candidate_answers = self.trainData['candidates']
X_train, Y_train = self.makeXY(self.trainData['questions'])
X_val, Y_val_true = self.makeXY(self.valData['questions'])
with open('/content/drive/My Drive/NLPA_Project/val_true.txt','wb') as f:
pickle.dump(Y_val_true,f)
#featurization
X_features_train, X_features_val = self.featurizerInstance.getFeatureRepresentation(X_train, X_val)
print(np.shape(X_features_train), np.shape(X_features_val))
self.clf = self.classifierInstance.buildClassifier(X_features_train, Y_train)
#Prediction
Y_val_pred = self.clf.predict(X_features_val)
print(Y_val_pred)
with open('/content/drive/My Drive/NLPA_Project/val_predict.txt','wb') as f:
pickle.dump(Y_val_pred,f)
self.evaluatorInstance = Evaluator()
a = self.evaluatorInstance.getAccuracy(Y_val_true, Y_val_pred)
p,r,f = self.evaluatorInstance.getPRF(Y_val_true, Y_val_pred)
print("Accuracy: " + str(a))
print("Precision: " + str(p))
print("Recall: " + str(r))
print("F-measure: " + str(f))
if __name__ == '__main__':
trainFilePath ='./data/train_formatted.json'
valFilePath = './data/test_formatted.json'
retrievalInstance = Retrieval()
#classifierInstance = MultinomialNaiveBayes()
featurizerInstance = CountFeaturizer()
#featurizerInstance = TfidfFeaturizer()()
classifierInstance = MLP()
trainInstance = Pipeline(trainFilePath, valFilePath, retrievalInstance, featurizerInstance, classifierInstance)