def tweetscore(sentence):
features = feature_extract.dialogue_act_features(sentence, topic_mod)
features_vec = vec.transform(features)
score = classifier.decision_function(features_vec)[0]
percentage = int(round(2.0 * (1.0 / (1.0 + np.exp(-score)) - 0.5) * 100.0))
return percentage
def tweetscore(sentence):
features = feature_extract.dialogue_act_features(sentence,topic_mod)
features_vec = vec.transform(features)
score = classifier.decision_function(features_vec)[0]
percentage = int(round(2.0*(1.0/(1.0+np.exp(-score))-0.5)*100.0))
return percentage
print 'Training topics'
topic_mod = topic.topic(nbtopic=200,alpha='symmetric')
topic_mod.fit(np.concatenate((pos_data,neg_data)))
print 'Feature eng'
# label set
cls_set = ['Non-Sarcastic','Sarcastic']
featuresets = []
index=0
for tweet in pos_data:
if (np.mod(index,10000)==0):
print "Positive tweet processed: ",index
featuresets.append((feature_extract.dialogue_act_features(tweet,topic_mod),cls_set[1]))
index+=1
index=0
for tweet in neg_data:
if (np.mod(index,10000)==0):
print "Negative tweet processed: ",index
featuresets.append((feature_extract.dialogue_act_features(tweet,topic_mod),cls_set[0]))
index+=1
featuresets=np.array(featuresets)
targets=(featuresets[0::,1]=='Sarcastic').astype(int)
print 'Dictionnary vectorizer'
vec = DictVectorizer()
featurevec = vec.fit_transform(featuresets[0::,0])
from sklearn.metrics import classification_report
from sklearn.feature_extraction import DictVectorizer
import pickle
import feature_extract
import heapq
pos_data = np.load('posproc.npy')
neg_data = np.load('negproc.npy')
class_set = ['Non-Sarcastic', 'Sarcastic']
featuresets = []
index = 0
for tweet in pos_data:
if (np.mod(index, 10000) == 0):
print "Positive tweet processed: ", index
featuresets.append((feature_extract.dialogue_act_features(tweet), class_set[1]))
index += 1
index = 0
for tweet in neg_data:
if (np.mod(index, 10000) == 0):
print "Negative tweet processed: ", index
featuresets.append((feature_extract.dialogue_act_features(tweet), class_set[0]))
index += 1
featuresets = np.array(featuresets)
targets = (featuresets[0::, 1] == 'Sarcastic').astype(int)
vec = DictVectorizer()
featurevec = vec.fit_transform(featuresets[0::, 0])
file_Name = "vecdict.p"
print('Training topics')
topic_mod = topic.topic(nbtopic=200, alpha='symmetric')
topic_mod.fit(np.concatenate((pos_data, neg_data)))
print('Feature eng')
# label set
cls_set = ['Non-Sarcastic', 'Sarcastic']
featuresets = []
index = 0
for tweet in pos_data:
if (np.mod(index, 10000) == 0):
print("Positive tweet processed: ", index)
featuresets.append(
(feature_extract.dialogue_act_features(tweet, topic_mod), cls_set[1]))
index += 1
index = 0
for tweet in neg_data:
if (np.mod(index, 10000) == 0):
print("Negative tweet processed: ", index)
featuresets.append(
(feature_extract.dialogue_act_features(tweet, topic_mod), cls_set[0]))
index += 1
featuresets = np.array(featuresets)
targets = (featuresets[0::, 1] == 'Sarcastic').astype(int)
print('Dictionnary vectorizer')
vec = DictVectorizer()
print 'Training topics'
topic_mod = topic.topic(nbtopic=200,alpha='symmetric')
topic_mod.fit(np.concatenate((pos_data,neg_data)))
print 'Feature eng'
# label set
cls_set = ['Non-Sarcastic','Sarcastic']
featuresets = []
index=0
for tweet in pos_data:
if (np.mod(index,10000)==0):
print "Positive tweet processed: ",index
featuresets.append((feature_extract.dialogue_act_features(tweet,topic_mod),cls_set[1]))
index+=1
index=0
for tweet in neg_data:
if (np.mod(index,10000)==0):
print "Negative tweet processed: ",index
featuresets.append((feature_extract.dialogue_act_features(tweet,topic_mod),cls_set[0]))
index+=1
featuresets=np.array(featuresets)
targets=(featuresets[0::,1]=='Sarcastic').astype(int)
print 'Dictionnary vectorizer'
vec = DictVectorizer()
featurevec = vec.fit_transform(featuresets[0::,0])
print 'Before Pickling'
pkl_file_1 = open('vecdict.p')
vec = pickle.load(pkl_file_1)
pkl_file = open('classif.p', 'rb')
classifier = pickle.load(pkl_file)
basic_test = []
file_test = open('test_MLWARE1.csv', 'r')
file_test_read = csv.reader(file_test)
i = 0
for each_tweet in file_test_read:
if i != 0:
basic_test.append(each_tweet[1])
else:
i = 1
feature_basictest = []
for tweet in basic_test:
feature_basictest.append(
feature_extract.dialogue_act_features(tweet, topic_mod))
feature_basictest = np.array(feature_basictest)
feature_basictestvec = vec.transform(feature_basictest)
print classifier.predict(feature_basictestvec)
file_test_open = open('test', 'w')
# file_test_write = csv.writer(file_test_open)
output = list(classifier.predict(feature_basictestvec))
for each_output in output:
file_test_open.write(str(each_output))
file_test_open.write('\n')
# file_test_write.writerow(each_output)
vec_file = open("vecdict.p", "rb")
vec = pickle.load(vec_file)
vec_file.close()
classifier_file = open("classif.p", "rb")
classifier = pickle.load(classifier_file)
classifier_file.close()
topic_mod_file = open("topic_mod.p", "rb")
topic_mod = pickle.load(topic_mod_file)
topic_mod_file.close()
#BASIC TEST
while (True):
print("Enter the text to test: ")
text = input()
features = []
features.append(feature_extract.dialogue_act_features(text, topic_mod))
features = np.array(features)
feature_vec = vec.transform(features)
ans = classifier.predict(feature_vec)
if ans == 1:
print("The text is sarcastic")
else:
print("The text is non-sarcastic")
print("The score of the text is: " +
str(classifier.decision_function(feature_vec)))
