-
Notifications
You must be signed in to change notification settings - Fork 0
/
myClassifier.py
317 lines (237 loc) · 10.9 KB
/
myClassifier.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
#!/usr/bin/env python3
#Code to apply the classifier on the blinded test data (runs on tira)
import pickle
import numpy as np
import json
from keras.utils import np_utils
from keras.preprocessing.sequence import pad_sequences
from datetime import datetime
from keras.models import model_from_yaml
import argparse
import os
modelPath = "/home/whitebait/models/"
outFileName ="results.jsonl"
parser = argparse.ArgumentParser(description='Apply clickbait model to file')
parser.add_argument('-i', action="store", help='Input Directory', type=str)
parser.add_argument('-o', action="store", help='Output Directory', type=str)
args = parser.parse_args()
if args.i == None or args.o == None:
print(parser.print_help())
exit()
inDir = args.i
outDir= args.o
inFile = os.path.join(inDir, 'instances.jsonl')
outFile = os.path.join(outDir, 'results.jsonl')
print("Reading instances from ='" +inFile +"'")
print("Writing result to ='" +outFile +"'")
### Fields in instances.jsonl:
class Data:
def __init__(self, id=None, postTimestamp=None, postText=None, postMedia=None, targetTitle=None, targetDescription=None, targetKeywords=None, targetParagraphs=None, targetCaptions=None, groundTruth=None):
self._id = id #"<instance id>",
self._postTimestamp = postTimestamp #"<weekday> <month> <day> <hour>:<minute>:<second> <time_offset> <year>",
self._postText = postText # ["<text of the post with links removed>"],
self._postMedia = postMedia #["<path to a file in the media archive>"],
self._targetTitle = targetTitle # <title of target article>",
self._targetDescription = targetDescription #"<description tag of target article>",
self._targetKeywords = targetKeywords # "<keywords tag of target article>",
self._targetParagraphs = targetParagraphs #["<text of the ith paragraph in the target article>"],
self._targetCaptions = targetCaptions #["<caption of the ith image in the target article>"]
self._groundTruth = groundTruth # Groundtruth, when applicable and provided in second file
@property
def id(self):
return self._id
@property
def postTimestamp(self):
return self._postTimestamp
@property
def postText(self):
return self._postText
@property
def postMedia(self):
return self._postMedia
@property
def targetTitle(self):
return self._targetTitle
@property
def targetDescription(self):
return self._targetDescription
@property
def targetKeywords(self):
return self._targetKeywords
@property
def targetParagraphs(self):
return self._targetParagraphs
@property
def targetCaptions(self):
return self._targetCaptions
@property
def groundTruth(self):
return self._groundTruth
@id.setter
def id(self, value):
self._id = value
@postTimestamp.setter
def postTimestamp(self, value):
self._postTimestamp = value
@postText.setter
def postText(self, value):
self._postText = value
@postMedia.setter
def postMedia(self, value):
self._postMedia = value
@targetTitle.setter
def targetTitle(self, value):
self._targetTitle = value
@targetDescription.setter
def targetDescription(self, value):
self._targetDescription = value
@targetKeywords.setter
def targetKeywords(self, value):
self._targetKeywords = value
@targetParagraphs.setter
def targetParagraphs(self, value):
self._targetParagraphs = value
@targetCaptions.setter
def targetCaptions(self, value):
self._targetCaptions = value
@groundTruth.setter
def groundTruth(self, value):
self._groundTruth = value
class GroundTruth:
def __init__(self, id=None, truthJudgments =None, truthMean =None, truthMedian=None, truthMode=None, truthClass=None):
self._id = id # "<instance id>",
self._truthJudgments = truthJudgments # [<number in [0,1]>],
self._truthMean = truthMean # <number in [0,1]>,
self._truthMedian = truthMedian # <number in [0,1]>,
self._truthMode = truthMode # <number in [0,1]>,
self._truthClass = truthClass # "clickbait | no-clickbait"
@property
def id(self):
return self._id
@property
def truthJudgments(self):
return self._truthJudgments
@property
def truthMean(self):
return self._truthMean
@property
def truthMedian(self):
return self._truthMedian
@property
def truthMode(self):
return self._truthMode
@property
def truthClass(self):
return self._truthClass
@id.setter
def id(self, value):
self._id = value
@truthJudgments.setter
def truthJudgments(self, value):
self._truthJudgments = value
@truthMean.setter
def truthMean(self, value):
self._truthMean = value
@truthMedian.setter
def truthMedian(self, value):
self._truthMedian = value
@truthMode.setter
def truthMode(self, value):
self._truthMode = value
@truthClass.setter
def truthClass(self, value):
self._truthClass = value
def parseData( line ):
inst = json.loads(line)
instance = Data(id=inst['id'], postTimestamp=inst['postTimestamp'], postText=inst['postText'], postMedia=inst['postMedia'], targetTitle=inst['targetTitle'],
targetDescription=inst['targetDescription'], targetKeywords=inst['targetKeywords'], targetParagraphs=inst['targetParagraphs'], targetCaptions=inst['targetCaptions'])
return instance
def parseGT( line ):
inst = json.loads(line)
instance = GroundTruth(id=inst['id'], truthJudgments =inst['truthJudgments'], truthMean =inst['truthMean'], truthMedian=inst['truthMedian'], truthMode=inst['truthMode'], truthClass=inst['truthClass'])
return instance
#1.) Load relevant processing data
file = open(modelPath +"processors.obj",'rb')
postTextTokenizer, targetTitleTokenizer, targetDescriptionTokenizer, targetKeywordsTokenizer, targetParagraphTokenizer, dayEncoder, hourEncoder, labelEncoder, classes, trainMedian, colnames, trainMean = pickle.load(file)
file = open(modelPath +"vars.obj",'rb')
MAX_POST_TEXT_LENGTH, MAX_TARGET_TITLE_LENGTH, MAX_TARGET_DESCRIPTION_LENGTH, MAX_TARGET_KEYWORDS_LENGTH, MAX_TARGET_PARAGRAPH_LENGTH = pickle.load(file)
# load YAML and create model
print("Loading yaml...")
yaml_file = open(modelPath +'modelReg.yaml', 'r')
loaded_model_yaml = yaml_file.read()
yaml_file.close()
final_model = model_from_yaml(loaded_model_yaml)
print("Loading weights")
final_model.load_weights(modelPath +"weightReg.h5")
print(final_model.summary())
instancesMap = {}
with open(inFile,'rb') as file:
for line in file:
instance = parseData(line.decode('utf-8'))
instancesMap[instance._id] = instance
###Extract relevant parts and store in list...
testIDs = []
testPostText = [] # "text of the post with links removed (e.g., The 15-year-old has been detained for 21 months.)
testPostMedia = [] # path to a file in the media archive (Single?)
testTitle = [] #"<title of target article>", (Single?)
testDescr = [] #"<description tag of target article>",
testKeywords = [] #"<keywords tag of target article>",
testParagraphs = [] # ["<text of the ith paragraph in the target article>"],
testCaptions = [] #["<caption of the ith image in the target article>"]
testTime = [] #"<weekday> <month> <day> <hour>:<minute>:<second> <time_offset> <year>",
for key in instancesMap:
testIDs.append(str(instancesMap[key]._id))
testPostText.append(str(instancesMap[key]._postText))
testPostMedia.append(str(instancesMap[key]._postMedia))
testTitle.append(str(instancesMap[key]._targetTitle))
testDescr.append(str(instancesMap[key]._targetDescription))
testKeywords.append(str(instancesMap[key]._targetKeywords))
testParagraphs.append(str(instancesMap[key]._targetParagraphs))
testCaptions.append(str(instancesMap[key]._targetCaptions))
testTime.append(datetime.strptime(instancesMap[key]._postTimestamp,
'%a %b %d %H:%M:%S +0000 %Y')) # %a - Weekday; %b month name; %d day; %H Hour (24-hour clock): %M
#############################
##Predict using model model
def predictToFile(predictions, predictToFile):
out_file = open(predictToFile, "w")
for i in range(predictions.shape[0]):
id = testIDs[i]
predValue = float(predictions[i])
predValue = max(0, predValue)
predValue = min(1, predValue)
my_dict = {
'id': id,
'clickbaitScore': predValue,
}
#print(predValue)
# print(placeName +" " +instance.text)
json.dump(my_dict, out_file)
out_file.write("\n")
out_file.close()
#1.) postTextModel
postTextSequence = postTextTokenizer.texts_to_sequences(testPostText)
postTextSequence = np.asarray(postTextSequence) # Convert to ndArray
postTextSequence = pad_sequences(postTextSequence, maxlen=MAX_POST_TEXT_LENGTH)
#2.) targetTitle
titleSequence= targetTitleTokenizer.texts_to_sequences(testTitle)
titleSequence = np.asarray(titleSequence) # Convert to ndArray
titleSequence = pad_sequences(titleSequence, maxlen=MAX_TARGET_TITLE_LENGTH)
#3.) targetDescription
descriptionSequence= targetDescriptionTokenizer.texts_to_sequences(testTitle)
descriptionSequence = np.asarray(descriptionSequence) # Convert to ndArray
descriptionSequence = pad_sequences(descriptionSequence, maxlen=MAX_TARGET_DESCRIPTION_LENGTH)
#4.) trainKeywords
keywordsSequence= targetKeywordsTokenizer.texts_to_sequences(testKeywords)
keywordsSequence = np.asarray(keywordsSequence) # Convert to ndArray
keywordsSequence = pad_sequences(keywordsSequence, maxlen=MAX_TARGET_KEYWORDS_LENGTH)
#5.) trainParagraphs
paragraphSequence= targetParagraphTokenizer.texts_to_sequences(testParagraphs)
paragraphSequence = np.asarray(paragraphSequence) # Convert to ndArray
paragraphSequence = pad_sequences(paragraphSequence, maxlen=MAX_TARGET_PARAGRAPH_LENGTH)
#7.) hour
testHour = list(map(lambda x: str(x.hour), testTime))
testHour = hourEncoder.transform(testHour)
testHour = np_utils.to_categorical(testHour)
##Final Model
predict = final_model.predict([postTextSequence, titleSequence, descriptionSequence, keywordsSequence, paragraphSequence, testHour])
predictToFile(predict, predictToFile=outFile)