def search():
negTweets = []
posTweets = []
with open('positiveTweets.csv', 'r') as csv_file:
csv_reader = csv.reader(csv_file)
for l in csv_reader:
posTweets.append([l[0], l[1]])
#tweets_ = tweets.tweets
with open('negativeTweets.csv', 'r') as csv_file:
csv_reader = csv.reader(csv_file)
for l in csv_reader:
negTweets.append([l[0], l[1]])
#tweets_ = tweets.tweets
# load absolute path of word lists
positives = os.path.join(sys.path[0], "positive-words.txt")
negatives = os.path.join(sys.path[0], "negative-words.txt")
# instantiate analyzer
analyzer = Analyzer(positives, negatives)
positive, negative, neutral = 0.0, 0.0, 0.0
cutoff = 0
if (len(negTweets) > len(posTweets)):
cutoff = len(posTweets) * 4 / 5
else:
cutoff = len(negTweets) * 4 / 5
tweets_ = negTweets[:cutoff] + posTweets[:cutoff]
testTweets_ = negTweets[cutoff:] + posTweets[cutoff:]
refsets = {'pos': set([]), 'neg': set([])}
testsets = {'pos': set([]), 'neg': set([])}
for tweet in tweets_:
score = analyzer.analyze(tweet[0])
if tweet[1] == 'pos':
testsets['pos'].add(tweet[0])
else:
testsets['neg'].add(tweet[0])
if score > 0.0:
positive += 1.0
refsets['pos'].add(tweet[0])
elif score <= 0.0:
negative += 1.0
refsets['neg'].add(tweet[0])
else:
neutral += 1.0
print('pos precision:', nltk.precision(refsets['pos'], testsets['pos']))
print('pos recall:', nltk.recall(refsets['pos'], testsets['pos']))
print('pos F-measure:', nltk.f_measure(refsets['pos'], testsets['pos']))
print('neg precision:', nltk.precision(refsets['neg'], testsets['neg']))
print('neg recall:', nltk.recall(refsets['neg'], testsets['neg']))
print('neg F-measure:', nltk.f_measure(refsets['neg'], testsets['neg']))
print(str(positive) + ',' + str(negative))
def evaluate_classifier(featx):
negids = movie_reviews.fileids('neg')
posids = movie_reviews.fileids('pos')
random.shuffle(negids)
random.shuffle(posids)
negfeats = [(featx(movie_reviews.words(fileids=[f])), 'neg') for f in negids]
posfeats = [(featx(movie_reviews.words(fileids=[f])), 'pos') for f in posids]
negcutoff = len(negfeats)*3/4
poscutoff = len(posfeats)*3/4
trainfeats = negfeats[:negcutoff] + posfeats[:poscutoff]
testfeats = negfeats[negcutoff:] + posfeats[poscutoff:]
classifier = NaiveBayesClassifier.train(trainfeats)
refsets = collections.defaultdict(set)
testsets = collections.defaultdict(set)
for i, (feats, label) in enumerate(testfeats):
refsets[label].add(i)
observed = classifier.classify(feats)
testsets[observed].add(i)
print 'accuracy:', nltk.classify.util.accuracy(classifier, testfeats)
print 'pos precision:', precision(refsets['pos'], testsets['pos'])
print 'pos recall:', recall(refsets['pos'], testsets['pos'])
print 'neg precision:', precision(refsets['neg'], testsets['neg'])
print 'neg recall:',recall(refsets['neg'], testsets['neg'])
classifier.show_most_informative_features()
def evaluate_naivebayes(classifier, test_reviews):
# For computing metrics
ref_set = collections.defaultdict(set)
test_set = collections.defaultdict(set)
ref_set_arr = []
test_set_arr = []
# Create gold standard and predicted labels
for i, (feat, label) in enumerate(test_reviews):
# Predict
observed = classifier.classify(feat)
ref_set[label].add(i)
test_set[observed].add(i)
label = 0 if label == "neg" else 1
observed = 0 if observed == "neg" else 1
ref_set_arr.append(label)
test_set_arr.append(observed)
print('pos precision:', precision(ref_set['pos'], test_set['pos']))
print('pos recall:', recall(ref_set['pos'], test_set['pos']))
print('neg precision:', precision(ref_set['neg'], test_set['neg']))
print('neg recall:', recall(ref_set['neg'], test_set['neg']))
print('misclassification rate', zero_one_loss(ref_set_arr, test_set_arr))
print('most informative features',
classifier.show_most_informative_features(10))
def evaluate_classifier(classifier, validationFeatures):
refsets = collections.defaultdict(set)
testsets = collections.defaultdict(set)
for i, (feats, label) in enumerate(validationFeatures):
refsets[label].add(i)
observed = classifier.classify(feats)
testsets[observed].add(i)
accuracy = nltk.classify.util.accuracy(classifier, validationFeatures)
pos_precision = precision(refsets['POSITIVE_TIME'], testsets['POSITIVE_TIME'])
pos_recall = recall(refsets['POSITIVE_TIME'], testsets['POSITIVE_TIME'])
pos_f_measure = f_measure(refsets['POSITIVE_TIME'], testsets['POSITIVE_TIME'])
neg_precision = precision(refsets['NEGATIVE_TIME'], testsets['NEGATIVE_TIME'])
neg_recall = recall(refsets['NEGATIVE_TIME'], testsets['NEGATIVE_TIME'])
neg_f_measure = f_measure(refsets['NEGATIVE_TIME'], testsets['NEGATIVE_TIME'])
print 'accuracy:', accuracy
print 'pos precision:', pos_precision
print 'pos recall:', pos_recall
print 'pos f-measure', pos_f_measure
print 'neg precision:', neg_precision
print 'neg recall:', neg_recall
print 'neg f-measure', neg_f_measure
return {'accuracy': accuracy, 'pos precision': pos_precision, 'pos recall': pos_recall, 'pos f-measure': pos_f_measure, 'neg precision': neg_precision, 'neg recall': neg_recall, 'neg f-measure': neg_f_measure}
def evaluation2(test_data, classifier):
refsets = collections.defaultdict(set)
testsets = collections.defaultdict(set)
for i, (feats, label) in enumerate(test_data):
refsets[label].add(i)
observed = classifier.classify(feats)
testsets[observed].add(i)
pos_precision = nltk.precision(refsets['Positive'], testsets['Positive'])
pos_recall = nltk.recall(refsets['Positive'], testsets['Positive'])
neg_precision = nltk.precision(refsets['Negative'], testsets['Negative'])
neg_recall = nltk.recall(refsets['Negative'], testsets['Negative'])
precision = (pos_precision + neg_precision) / 2
recall = (pos_recall + neg_recall) / 2
try: #Try loop added for neutrals
neu_precision = nltk.precision(refsets['Neutral'], testsets['Neutral'])
neu_recall = nltk.recall(refsets['Neutral'], testsets['Neutral'])
precision = (pos_precision + neg_precision + neu_precision) / 3
recall = (pos_recall + neg_recall + neu_recall) / 3
except:
pass
return precision, recall
def evaluate_classifier(featx):
negids = nltk.corpus.movie_reviews.fileids('neg')
posids = nltk.corpus.movie_reviews.fileids('pos')
negfeats = [(featx(nltk.corpus.movie_reviews.words(fileids=[f])), 'neg') for f in negids]
posfeats = [(featx(nltk.corpus.movie_reviews.words(fileids=[f])), 'pos') for f in posids]
negcutoff = int(len(negfeats) * 3 / 4)
poscutoff = int(len(posfeats) * 3 / 4)
trainfeats = negfeats[:negcutoff] + posfeats[:poscutoff]
testfeats = negfeats[negcutoff:] + posfeats[poscutoff:]
print('train on %d instances, test on %d instances' % (len(trainfeats), len(testfeats)))
classifier = nltk.classify.NaiveBayesClassifier.train(trainfeats)
refsets = collections.defaultdict(set)
testsets = collections.defaultdict(set)
for i, (feats, label) in enumerate(testfeats):
refsets[label].add(i)
observed = classifier.classify(feats)
testsets[observed].add(i)
print('pos precision', nltk.precision(refsets['pos'], testsets['pos']))
print('pos recall', nltk.recall(refsets['pos'], testsets['pos']))
print('neg precision', nltk.precision(refsets['neg'], testsets['neg']))
print('neg recall', nltk.recall(refsets['neg'], testsets['neg']))
print('accuracy:', nltk.classify.util.accuracy(classifier, testfeats))
classifier.show_most_informative_features()
def evaluate_features(feature_select):
posFeatures = []
negFeatures = []
#http://stackoverflow.com/questions/367155/splitting-a-string-into-words-and-punctuation
#breaks up the sentences into lists of individual words (as selected by the input mechanism) and appends 'pos' or 'neg' after each list
with open(RT_POLARITY_POS_FILE, "rb") as f:
posSentences = f.readlines()
#posSentences = pos_data.split('\n')
random.shuffle(posSentences)
with open(RT_POLARITY_NEG_FILE, "rb") as f:
#negSentences = f.read().split('\n')
negSentences = f.readlines()
random.shuffle(negSentences)
#with open(RT_POLARITY_POS_FILE, 'r') as posSentences:
for i in posSentences:
posWords = re.findall(r"[\w']+|[.,!?;@#]", str(i).rstrip())
posWords = [feature_select(posWords), 'pos'] #pos = contains location
posFeatures.append(posWords)
#with open(RT_POLARITY_NEG_FILE, 'r') as negSenBufferedReader: <_io.BufferedReader name='/home/ira/Dropbox/twitter/contain_location_tweets.txt'>tences:
for i in negSentences:
negWords = re.findall(r"[\w']+|[.,!?;@#]", str(i).rstrip())
negWords = [feature_select(negWords), 'neg'] #neg = doesn't contain location
negFeatures.append(negWords)
#selects 3/4 of the features to be used for training and 1/4 to be used for testing
posCutoff = int(math.floor(len(posFeatures)*3/4))
negCutoff = int(math.floor(len(negFeatures)*3/4))
trainFeatures = posFeatures[:posCutoff] + negFeatures[:negCutoff]
testFeatures = posFeatures[posCutoff:] + negFeatures[negCutoff:]
#trains a Naive Bayes Classifier
classifier = NaiveBayesClassifier.train(trainFeatures)
#initiates referenceSets and testSets
referenceSets = collections.defaultdict(set)
testSets = collections.defaultdict(set)
#puts correctly labeled sentences in referenceSets and the predictively labeled version in testsets
for i, (features, label) in enumerate(testFeatures):
referenceSets[label].add(i)
predicted = classifier.classify(features)
testSets[predicted].add(i)
#prints metrics to show how well the feature selection did
print ('train on %d instances, test on %d instances' % (len(trainFeatures), len(testFeatures)) )
print ('accuracy:', nltk.classify.util.accuracy(classifier, testFeatures))
#print 'pos precision:', nltk.metrics.precision(referenceSets['pos'], testSets['pos'])
print ('pos precision:', precision(referenceSets['pos'], testSets['pos']) )
print ('pos recall:', recall(referenceSets['pos'], testSets['pos']) )
#print 'neg precision:', nltk.metrics.precision(referenceSets['neg'], testSets['neg'])
print ('neg precision:',precision(referenceSets['neg'], testSets['neg']))
print ('neg recall:', recall(referenceSets['neg'], testSets['neg']))
classifier.show_most_informative_features(10)
def evaluate(train_set, test_set, classifier, name): refsets = collections.defaultdict(set) testsets = collections.defaultdict(set) for i, (features, label) in enumerate(test_set): refsets[label].add(i) observed = classifier.classify(features) testsets[observed].add(i) # Get accuracy on training set, test set and get positive and negative recall. trainacc = 100 * classify.accuracy(classifier, train_set) testacc = 100 * classify.accuracy(classifier, test_set) spam_false = 100 - nltk.recall(refsets['spam'], testsets['spam'])*100 ham_false = 100 - nltk.recall(refsets['ham'], testsets['ham'])*100 return trainacc, testacc, spam_false, ham_false
def train(feature_name, feature_detector, use_best_words, limit):
# call function read the text(21) and split into sentiment(23) return a best words by limit
neudata, posdata, negdata, alldata, best_words = split_sets(
selected_input, 21, 23, limit)
# only some of the tokinator help function uses the best_words input which is handled by the
# use_best_words parameter
negfeats = split_feats('Negative', negdata, feature_detector,
use_best_words, best_words)
posfeats = split_feats('Positive', posdata, feature_detector,
use_best_words, best_words)
neufeats = split_feats('Neutral', neudata, feature_detector,
use_best_words, best_words)
trainfeats, testfeats = create_train_test_sets(negfeats, posfeats,
neufeats)
classifier = NaiveBayesClassifier.train(trainfeats)
refsets = collections.defaultdict(set)
testsets = collections.defaultdict(set)
for i, (feats, label) in enumerate(testfeats):
refsets[label].add(i)
observed = classifier.classify(feats)
testsets[observed].add(i)
accuracy = nltk.classify.util.accuracy(classifier, testfeats)
pos_precision = precision(refsets['Positive'], testsets['Positive'])
pos_recall = recall(refsets['Positive'], testsets['Positive'])
neg_precision = precision(refsets['Negative'], testsets['Negative'])
neg_recall = recall(refsets['Negative'], testsets['Negative'])
neu_precision = precision(refsets['Neutral'], testsets['Neutral'])
neu_recall = recall(refsets['Neutral'], testsets['Neutral'])
classifier_with_accuracy = {
'classifier': classifier,
'feature_name': feature_name,
'feature_detector': feature_detector,
'best_words': use_best_words,
'accuracy': accuracy,
'limit': limit,
'best_words_list': best_words,
'pos_precision': pos_precision,
'pos_recall': pos_recall,
'neg_precision': neg_precision,
'neg_recall': neg_recall,
'neu_precision': neu_precision,
'neu_recall': neu_recall
}
return classifier_with_accuracy
def KFoldAccuracy(all_reviews):
global word_features
from sklearn.model_selection import KFold
import numpy as np
import collections
k_fold = KFold(n_splits=10, random_state=1992, shuffle=True)
featuresets = np.array(all_reviews)
accuracy_scores = []
for train_set, test_set in k_fold.split(featuresets):
word_features = get_word_features(
get_words_in_reviews(featuresets[train_set].tolist()))
train_features = nltk.classify.apply_features(
extract_features, featuresets[train_set].tolist())
test_features = nltk.classify.apply_features(
extract_features, featuresets[test_set].tolist())
classifier = nltk.NaiveBayesClassifier.train(train_features)
refsets = collections.defaultdict(set)
testsets = collections.defaultdict(set)
for i, (feats, label) in enumerate(test_features):
refsets[label].add(i)
observed = classifier.classify(feats)
testsets[observed].add(i)
accuracy_scores.append(
nltk.classify.util.accuracy(classifier, test_features))
# f1_pos = nltk.f_measure(refsets['pos'], testsets['pos'])
f1_neg = nltk.f_measure(refsets['neg'], testsets['neg'])
f1_swear = nltk.f_measure(refsets['swear'], testsets['swear'])
# pre_pos = nltk.precision(refsets['pos'], testsets['pos'])
pre_neg = nltk.precision(refsets['neg'], testsets['neg'])
pre_swear = nltk.precision(refsets['swear'], testsets['swear'])
# re_pos = nltk.recall(refsets['pos'], testsets['pos'])
re_neg = nltk.recall(refsets['neg'], testsets['neg'])
re_swear = nltk.recall(refsets['swear'], testsets['swear'])
print(f'train: {len(train_set)} test: {len(test_set)}')
print('=================== Results ===================')
print(f'Accuracy {accuracy_scores[-1]:f}')
print(' Negative Swear')
print(f'F1 {f1_neg:f} {f1_swear:f}]')
print(f'Precision {pre_neg:f} {pre_swear:f}]')
print(f'Recall {re_neg:f} {re_swear:f}]')
print('===============================================\n')
def check_mutato_accuracy(input_path):
with open(input_path, 'r', encoding=encoding) as csv_train:
csv_reader = csv.reader(csv_train, delimiter=delim)
refsets = collections.defaultdict(set)
testsets = collections.defaultdict(set)
results = []
gold = []
label = ''
for i, row in enumerate(csv_reader):
if row[23] == '2':
label = 'Negative'
elif row[23] == '1':
label = 'Positive'
elif row[23] == '0':
label = 'Neutral'
gold.append(label)
refsets[label].add(i)
testsets[row[26]].add(i)
results.append(row[26])
# Since we dont have a classifier object we manually calculate the accuracy
# We do it the same way as nltk.util.accuracy but instead of using the classifier to compare
# We just compare between the result and the gold standard (refset)
correct = [l == r for (l, r) in zip(gold, results)]
accuracy_manual = sum(correct) / len(correct)
pos_precision = precision(refsets['Positive'], testsets['Positive'])
pos_recall = recall(refsets['Positive'], testsets['Positive'])
neg_precision = precision(refsets['Negative'], testsets['Negative'])
neg_recall = recall(refsets['Negative'], testsets['Negative'])
neu_precision = precision(refsets['Neutral'], testsets['Neutral'])
neu_recall = recall(refsets['Neutral'], testsets['Neutral'])
classifier_with_accuracy = {
'classifier': 'N/A',
'feature_name': 'MUTATO',
'feature_detector': 'MUTATO',
'best_words': False,
'accuracy': accuracy_manual,
'limit': 0,
'pos_precision': pos_precision,
'pos_recall': pos_recall,
'neg_precision': neg_precision,
'neg_recall': neg_recall,
'neu_precision': neu_precision,
'neu_recall': neu_recall
}
return classifier_with_accuracy
def evaluate_classifier(featx):
reviews = product_reviews_1.reviews()
reviewlines = []
for review in reviews:
for line in review.review_lines:
reviewlines.append(line)
sentfeats = [[(featx(line.sent), f[1][0]) for f in line.features] for line in reviewlines if len(line.features)>0]
plusfeats = []
minusfeats = []
for sentfeat in sentfeats:
for feat in sentfeat:
if feat[1] == "+":
plusfeats.append(feat)
elif feat[1] == "-":
minusfeats.append(feat)
if len(minusfeats) > len(plusfeats):
minusfeats = minusfeats[:len(plusfeats)]
else:
plusfeats = plusfeats[:len(minusfeats)]
minuscutoff = int(len(minusfeats)*3/4)
pluscutoff = int(len(plusfeats)*3/4)
trainfeats = minusfeats[:minuscutoff] + plusfeats[:pluscutoff]
testfeats = minusfeats[minuscutoff:] + plusfeats[pluscutoff:]
print('train on %d instances, test on %d instances' % (len(trainfeats), len(testfeats)))
classifier = NaiveBayesClassifier.train(trainfeats)
refsets = collections.defaultdict(set)
testsets = collections.defaultdict(set)
for i, (feats, label) in enumerate(testfeats):
refsets[label].add(i)
observed = classifier.classify(feats)
testsets[observed].add(i)
#return nltk.classify.util.accuracy(classifier, testfeats)
print('accuracy:', nltk.classify.util.accuracy(classifier, testfeats))
print('pos precision:', precision(refsets['+'], testsets['+']))
print('pos recall:', recall(refsets['+'], testsets['+']))
print('neg precision:', precision(refsets['-'], testsets['-']))
print('neg recall:', recall(refsets['-'], testsets['-']))
classifier.show_most_informative_features()
print(classifier.classify(featx(["I", "hate", "it", "."])))
print(classifier.classify(featx(["I", "love", "it", "."])))
def print_precision_recall(classifier, test_set):
""" Computes and prints the precision and recall metrics, given a classifier
and a test set
"""
known_set = collections.defaultdict(set)
computed_set = collections.defaultdict(set)
for i, (features, label) in enumerate(test_set):
known_set[label].add(i)
predicted = classifier.classify(features)
computed_set[predicted].add(i)
print('pos precision:', precision(known_set['pos'], computed_set['pos']))
print('pos recall:', recall(known_set['pos'], computed_set['pos']))
print('neg precision:', precision(known_set['neg'], computed_set['neg']))
print('neg recall:', recall(known_set['neg'], computed_set['neg']))
def evaluation(self, test_set, classifier):
referenceSet = collections.defaultdict(set)
testSet = collections.defaultdict(set)
referenceSet_cm = []
testSet_cm = []
for index, (sentences, actualLabel) in enumerate(test_set):
referenceSet[actualLabel].add(index)
referenceSet_cm.append(actualLabel)
predictedLabel = classifier.classify(sentences)
testSet[predictedLabel].add(index)
testSet_cm.append(predictedLabel)
print referenceSet.keys()
print testSet.keys()
print("-------------Claim metrics-----------")
print('Accuracy of the classifier: ',
nltk.classify.util.accuracy(classifier, test_set))
print(
'precision: ',
precision(referenceSet[referenceSet.keys()[0]],
testSet[testSet.keys()[0]]))
print(
'recall: ',
recall(referenceSet[referenceSet.keys()[0]],
testSet[testSet.keys()[0]]))
print(
'F-measure: ',
f_measure(referenceSet[referenceSet.keys()[0]],
testSet[testSet.keys()[0]]))
print("-------------Premise metrics-----------")
print('Accuracy of the classifier: ',
nltk.classify.util.accuracy(classifier, test_set))
print(
'precision: ',
precision(referenceSet[referenceSet.keys()[1]],
testSet[testSet.keys()[1]]))
print(
'recall: ',
recall(referenceSet[referenceSet.keys()[1]],
testSet[testSet.keys()[1]]))
print(
'F-measure: ',
f_measure(referenceSet[referenceSet.keys()[1]],
testSet[testSet.keys()[1]]))
def evaluate_features(feature_select):
posFeatures = []
negFeatures = []
#http://stackoverflow.com/questions/367155/splitting-a-string-into-words-and-punctuation
#breaks up the sentences into lists of individual words (as selected by the input mechanism) and appends 'pos' or 'neg' after each list
with open('rt-polarity-pos.txt', 'r') as posSentences:
for i in posSentences:
posWords = re.findall(r"[\w']+|[.,!?;]", i.rstrip())
posWords = [feature_select(posWords), 'pos']
posFeatures.append(posWords)
with open('rt-polarity-neg.txt', 'r') as negSentences:
for i in negSentences:
negWords = re.findall(r"[\w']+|[.,!?;]", i.rstrip())
negWords = [feature_select(negWords), 'neg']
negFeatures.append(negWords)
#selects 3/4 of the features to be used for training and 1/4 to be used for testing
posCutoff = int(math.floor(len(posFeatures) * 3 / 4))
negCutoff = int(math.floor(len(negFeatures) * 3 / 4))
trainFeatures = posFeatures[:posCutoff] + negFeatures[:negCutoff]
testFeatures = posFeatures[posCutoff:] + negFeatures[negCutoff:]
#trains a Naive Bayes Classifier
classifier = NaiveBayesClassifier.train(trainFeatures)
#initiates referenceSets and testSets
referenceSets = collections.defaultdict(set)
testSets = collections.defaultdict(set)
#puts correctly labeled sentences in referenceSets and the predictively labeled version in testsets
for i, (features, label) in enumerate(testFeatures):
referenceSets[label].add(i)
predicted = classifier.classify(features)
testSets[predicted].add(i)
#prints metrics to show how well the feature selection did
print('train on %d instances, test on %d instances' %
(len(trainFeatures), len(testFeatures)))
print('accuracy:', nltk.classify.util.accuracy(classifier, testFeatures))
print('pos precision:',
nltk.precision(referenceSets['pos'], testSets['pos']))
print('pos recall:', nltk.recall(referenceSets['pos'], testSets['pos']))
print('neg precision:',
nltk.precision(referenceSets['neg'], testSets['neg']))
print('neg recall:', nltk.recall(referenceSets['neg'], testSets['neg']))
classifier.show_most_informative_features(10)
def classifier_Train(self, train, test, labels):
print('Training the Classifier (%s Classifier)' % labels)
train_set_final = []
test_set_final = []
train_text = []
for item in train:
train_text.append((item[0] + item[1], item[2])) #review text,label
train_set_final.append((item[0] + item[1], item[2]))
for item in test:
test_set_final.append((item[0] + item[1], item[2]))
start_time = time.time()
self.word_features = _get_words_from_dataset(train_set_final)
train_features_labels = [(self.extractor(d), c)
for d, c in train_set_final]
cl = class_.train(train_features_labels, **class_params)
end_time = time.time()
P = collections.defaultdict(set)
R = collections.defaultdict(set)
F = collections.defaultdict(set)
test_features_labels = [(self.extractor(d), c)
for d, c in test_set_final]
for label in labels:
refsets = collections.defaultdict(set)
testsets = collections.defaultdict(set)
for i, (feats, lbl) in enumerate(test_features_labels):
refsets[lbl].add(i)
observed = cl.classify(feats)
testsets[observed].add(i)
p = nltk.precision(refsets[label], testsets[label])
if p is None:
p = 0
r = nltk.recall(refsets[label], testsets[label])
if r is None:
r = 0
f = nltk.f_measure(refsets[label], testsets[label])
if f is None:
f = 0
print "label: %s" % label
print("P:%s, R:%s, F:%s" % (p, r, f))
P[label] = p
R[label] = r
F[label] = f
cl.show_most_informative_features(10)
return P, R, F, (end_time - start_time)
def evaluation(test_data, classifier):
# accuracy=np.round(classify.accuracy(classifier, test_data),3)
# print("Accuracy of Model is:", accuracy)
# print(classifier.show_most_informative_features(10))
refsets = collections.defaultdict(set)
testsets = collections.defaultdict(set)
for i, (feats, label) in enumerate(test_data):
refsets[label].add(i)
observed = classifier.classify(feats)
testsets[observed].add(i)
pos_precision = np.round(
nltk.precision(refsets['Positive'], testsets['Positive']), 3)
pos_recall = np.round(
nltk.recall(refsets['Positive'], testsets['Positive']), 3)
pos_fscore = np.round(
nltk.f_measure(refsets['Positive'], testsets['Positive']), 3)
neg_precision = np.round(
nltk.precision(refsets['Negative'], testsets['Negative']), 3)
neg_recall = np.round(
nltk.recall(refsets['Negative'], testsets['Negative']), 3)
neg_fscore = np.round(
nltk.f_measure(refsets['Negative'], testsets['Negative']), 3)
print('pos precision:', pos_precision)
print('pos recall:', pos_recall)
print('pos F-score:', pos_fscore)
print('neg precision:', neg_precision)
print('neg recall:', neg_recall)
print('neg F-score:', neg_fscore)
try: # Try loop added for neutrals
neu_precision = np.round(
nltk.precision(refsets['Neutral'], testsets['Neutral']), 3)
neu_recall = np.round(
nltk.recall(refsets['Neutral'], testsets['Neutral']), 3)
neu_fscore = np.round(
nltk.f_measure(refsets['Neutral'], testsets['Neutral']), 3)
print('neu precision:', neu_precision)
print('neu recall:', neu_recall)
print('neu F-score:', neu_fscore)
except:
pass
def evaluate_classifier(featx):
#print(featx)
neg_dict = movie_words('neg')
pos_dict = movie_words('pos')
negfeats = []
posfeats = []
for word in featx:
try:
neg_dict[word]
negfeats.append(({word: True}, 'neg'))
except KeyError:
print(word + " Missing from negative")
try:
pos_dict[word]
posfeats.append(({word: True}, 'neg'))
except KeyError:
print(word + " Missing from positive")
negcutoff = len(negfeats)*3//4
poscutoff = len(posfeats)*3//4
print(negcutoff)
print(poscutoff)
trainfeats = negfeats[:negcutoff] + posfeats[:poscutoff]
testfeats = negfeats[negcutoff:] + posfeats[poscutoff:]
classifier = NaiveBayesClassifier.train(trainfeats)
refsets = collections.defaultdict(set)
testsets = collections.defaultdict(set)
for i, (feats, label) in enumerate(testfeats):
refsets[label].add(i)
observed = classifier.classify(feats)
testsets[observed].add(i)
print('accuracy:', nltk.classify.util.accuracy(classifier, testfeats))
print('pos precision:', precision(refsets['pos'], testsets['pos']))
print('pos recall:', recall(refsets['pos'], testsets['pos']))
print('neg precision:', precision(refsets['neg'], testsets['neg']))
print('neg recall:', recall(refsets['neg'], testsets['neg']))
classifier.show_most_informative_features()
def __init__(self):
#reading pre-labeled input and splitting into lines
posSentences = open('negMessages.txt', 'r')
negSentences = open('posMessages.txt', 'r')
posSentences = re.split(r'\n', posSentences.read())
negSentences = re.split(r'\n', negSentences.read())
posFeatures = []
negFeatures = []
#breaks up the sentences into lists of individual words (as selected by the input mechanism) and appends 'pos' or 'neg' after each list
for i in posSentences:
posWords = re.findall(r"[\w']+|[.,!?;]", i)
posWords = [self.make_full_dict(posWords), 'pos']
posFeatures.append(posWords)
for i in negSentences:
negWords = re.findall(r"[\w']+|[.,!?;]", i)
negWords = [self.make_full_dict(negWords), 'neg']
negFeatures.append(negWords)
# selects 3/4 of the features to be used for training and 1/4 to be used for testing
posCutoff = int(math.floor(len(posFeatures) * 3 / 4))
negCutoff = int(math.floor(len(negFeatures) * 3 / 4))
trainFeatures = posFeatures[:posCutoff] + negFeatures[:negCutoff]
testFeatures = posFeatures[posCutoff:] + negFeatures[negCutoff:]
self.classifier = NaiveBayesClassifier.train(trainFeatures)
referenceSets = {'pos':set([]), 'neg':set([])}
testSets = {'pos':set([]), 'neg':set([])}
for i, (features, label) in enumerate(testFeatures):
referenceSets[label].add(i)
predicted = self.classifier.classify(features)
testSets[predicted].add(i)
posPrecision = precision(reference=referenceSets['pos'], test=testSets['pos'])
print 'train on %d instances, test on %d instances' % (len(trainFeatures), len(testFeatures))
print 'accuracy:', nltk.classify.util.accuracy(self.classifier, testFeatures)
print 'pos precision:', precision(referenceSets['pos'], testSets['pos'])
print 'pos recall:', recall(referenceSets['pos'], testSets['pos'])
print 'neg precision:', precision(referenceSets['neg'], testSets['neg'])
print 'neg recall:',recall(referenceSets['neg'], testSets['neg'])
self.classifier.show_most_informative_features(10)
def predict(self, data_vectorized):
start_time = time.time()
# format the testfeats in the format nltk asks for
# use the word without vectorizing
testfeats = []
logging.info('create the testing feats')
for i, feat in enumerate(data_vectorized['x_test']):
feat_as_words = self.bigram_word_feats(
self.tokenizer.tokenize(feat))
label = data_vectorized['y_test'][i]
testfeats.append((feat_as_words, label))
refsets = collections.defaultdict(set)
testsets = collections.defaultdict(set)
for i, (feats, label) in enumerate(testfeats):
refsets[label].add(i)
observed = self.classifier.classify(feats)
testsets[observed].add(i)
self.time_prediction = (time.time() - start_time)
logging.info("Results for" + self.name + "with nltk scoring methods")
logging.info("Training time: %fs; Prediction time: %fs" %
(self.time_training, self.time_prediction))
# logging.info(
# classification_report(data_vectorized['y_test'], self.predictions, target_names=target_names))
logging.info('--- accuracy: %6.2f ---' %
nltk.classify.util.accuracy(self.classifier, testfeats))
logging.info('--- pos precision: %6.2f ---' %
precision(refsets[1], testsets[1]))
logging.info('--- pos recall: %6.2f ---' %
recall(refsets[1], testsets[1]))
logging.info('--- neg precision: %6.2f ---' %
precision(refsets[0], testsets[0]))
logging.info('--- neg recall: %6.2f ---' %
recall(refsets[0], testsets[0]))
logging.info("--- testing done - %6.2f seconds ---" %
(time.time() - start_time))
logging.info(self.classifier.most_informative_features(n=10))
self.classifier.show_most_informative_features()
def performance(self, test_data):
prediction = self.predict(test_data)
pos_loc = set(
[i for i in range(len(prediction)) if prediction[i] == 'pos'])
neg_loc = set(range(len(prediction))) - pos_loc
pos_ref = set(
[i for i in range(len(prediction)) if test_data[i][1] == 'pos'])
neg_ref = set(range(len(prediction))) - pos_ref
print('===============================\n')
print('Model Summary:\n')
print(self.classifier_type + ' with features ' + self.feats.__name__ +
'\n')
print('Overall Accuracy: %.3f\n' %
(nltk.classify.util.accuracy(self.classifier, test_data)))
print('Positive Precision: %.3f\n' %
(nltk.precision(pos_ref, pos_loc)))
print('Positive Recall: %.3f\n' % (nltk.recall(pos_ref, pos_loc)))
print('Negative Precision: %.3f\n' %
(nltk.precision(neg_ref, neg_loc)))
print('Negative Recall: %.3f\n' % (nltk.recall(neg_ref, neg_loc)))
def classifier_metrics(self, test_set):
"""Method to print the classifier metrics(precision, recall, accuracy, f-measure) and the most informative features
The NLTK metrics module provides functions for calculating all three metrics but we need build 2 sets for each classification label
Args:
comment text
Returns:
None
"""
classifier = self.model
refsets = collections.defaultdict(set)
testsets = collections.defaultdict(set)
for i, (features, label) in enumerate(test_set):
refsets[label].add(i)
observed = classifier.classify(features)
testsets[observed].add(i)
print("The Naive bayes classifier accuracy is : {}".format(
nltk.classify.accuracy(classifier, test_set)))
print('')
print(
"The Naive bayes classifier positive sentiment precision is : {}".
format(precision(refsets['pos'], testsets['pos'])))
print(
"The Naive bayes classifier negative sentiment precision is : {}".
format(precision(refsets['neg'], testsets['neg'])))
print('')
print("The Naive bayes classifier positive sentiment recall is : {}".
format(recall(refsets['pos'], testsets['pos'])))
print("The Naive bayes classifier negative sentiment recall is : {}".
format(recall(refsets['neg'], testsets['neg'])))
print('')
print(
"The Naive bayes classifier positive sentiment f-measure is : {}".
format(f_measure(refsets['pos'], testsets['pos'])))
print(
"The Naive bayes classifier negative sentiment f-measure is : {}".
format(f_measure(refsets['neg'], testsets['neg'])))
print('')
print('classifier top 5 most informative features is {}'.format(
classifier.show_most_informative_features(5)))
def precision_recall (classifier, test_set): #pour faire listes des precissions et rappels
refsets=collections.defaultdict(set)
testset=collections.defaultdict(set)
for i, (sent, category) in enumerate(test_set):
refsets[category].add(i)
observed=classifier.classify(sent)
testset[observed].add(i)
prec={} #list pour preccision
rapp={} #list pour rappel
for category in leMonde.categories(): #pour chaque categorie
prec[category]=nltk.precision(refsets[category], testset[category]) #preccision pour cet categorie
rapp[category]=nltk.recall(refsets[category],testset[category]) #rappel pour cet categorie
return prec,rapp #returer ces listes
def precision_recall (classifier, test_set):
refsets=collections.defaultdict(set)
testset=collections.defaultdict(set)
for i, (sent, category) in enumerate(test_set):
refsets[category].add(i)
observed=classifier.classify(sent)
testset[observed].add(i)
prec={}
rec={}
for category in leMonde.categories():
prec[category]=nltk.precision(refsets[category], testset[category])
rec[category]=nltk.recall(refsets[category],testset[category])
return prec,rec
def precision_recall(classifier, test_set):
refsets = collections.defaultdict(set)
testsets = collections.defaultdict(set)
for i, (sent, category) in enumerate(test_set):
refsets[category].add(i)
observed = classifier.classify(sent)
testsets[observed].add(i)
prec = {}
rec = {}
for category in leMonde.categories():
prec[category] = nltk.precision(refsets[category], testsets[category])
rec[category] = nltk.recall(refsets[category], testsets[category])
return prec, rec
def log_precision_recall(classifier, test_set, file):
""" Computes and logs the precision and recall metrics into a file, given a
classifier and a test set
"""
known_set = collections.defaultdict(set)
computed_set = collections.defaultdict(set)
for i, (features, label) in enumerate(test_set):
known_set[label].add(i)
predicted = classifier.classify(features)
computed_set[predicted].add(i)
line = 'pos precision: ' + str(
precision(known_set['pos'], computed_set['pos']))
file.write(line + '\n')
line = 'pos recall: ' + str(recall(known_set['pos'], computed_set['pos']))
file.write(line + '\n')
line = 'neg precision: ' + str(
precision(known_set['neg'], computed_set['neg']))
file.write(line + '\n')
line = 'neg recall: ' + str(recall(known_set['neg'], computed_set['neg']))
file.write(line + '\n')
def get_accuracy_measures(classifier, testing_data, p_label):
actuallabels =collections.defaultdict(set)
predictedlabels = collections.defaultdict(set)
for i, (tokens, label) in enumerate(testing_data):
actuallabels[label].add(i)
predicted = classifier.classify(tokens)
predictedlabels[predicted].add(i)
result=[]
result.append(precision(actuallabels[p_label], predictedlabels[p_label]))
result.append(recall(actuallabels[p_label], predictedlabels[p_label]))
result.append(f_measure(actuallabels[p_label], predictedlabels[p_label]))
return result
def evaluate_svm(y_hat, test_reviews):
# For computing metrics
ref_set = collections.defaultdict(set)
test_set = collections.defaultdict(set)
ref_set_arr = []
test_set_arr = []
# Create gold standard and predicted labels
for i, label in enumerate(test_reviews):
label = "neg" if label == 0 else "pos"
observed = "neg" if y_hat[i] == 0 else "pos"
ref_set[label].add(i)
test_set[observed].add(i)
label = 0 if label == "neg" else 1
observed = 0 if observed == "neg" else 1
ref_set_arr.append(label)
test_set_arr.append(observed)
print('pos precision:', precision(ref_set['pos'], test_set['pos']))
print('pos recall:', recall(ref_set['pos'], test_set['pos']))
print('neg precision:', precision(ref_set['neg'], test_set['neg']))
print('neg recall:', recall(ref_set['neg'], test_set['neg']))
print('misclassification rate', zero_one_loss(ref_set_arr, test_set_arr))
def evaluate_model(MaxEntClassifier):
refsets = collections.defaultdict(set)
testsets = collections.defaultdict(set)
accuracy = classify.accuracy(MaxEntClassifier, validation_features)*100
accuracy_list.append(accuracy)
for i, (feats, label) in enumerate(validation_features):
refsets[label].add(i)
observed = MaxEntClassifier.classify(feats)
testsets[observed].add(i)
negative_precision = precision(refsets['negative'], testsets['negative'])
positive_precision = precision(refsets['positive'], testsets['positive'])
positive_recall = recall(refsets['positive'], testsets['positive'])
negative_recall = recall(refsets['negative'], testsets['negative'])
try:
avg_recall = 0.5*(negative_recall+positive_recall)
avg_precision = 0.5*(negative_precision+positive_precision)
precision_list.append(avg_precision)
recall_list.append(avg_recall)
except TypeError:
pass
return precision_list, recall_list, accuracy_list
def get_pr(classifier, features_train, features_test):
refsets = defaultdict(set)
testsets = defaultdict(set)
for i, (feats, label) in enumerate(features_test):
refsets[label].add(i)
observed = classifier.classify(feats)
testsets[observed].add(i)
# [0] = positive precision
# [1] = negative precision
p = [
precision(refsets['Positive'], testsets['Positive']),
precision(refsets['Negative'], testsets['Negative'])
]
# [0] = positive recall
# [1] = negative recall
r = [
recall(refsets['Positive'], testsets['Positive']),
recall(refsets['Negative'], testsets['Negative'])
]
return [p, r]
def evaluate_classifier(classifier, validationFeatures):
refsets = collections.defaultdict(set)
testsets = collections.defaultdict(set)
for i, (feats, label) in enumerate(validationFeatures):
refsets[label].add(i)
observed = classifier.classify(feats)
testsets[observed].add(i)
accuracy = nltk.classify.util.accuracy(classifier, validationFeatures)
pos_precision = precision(refsets['POSITIVE_TIME'],
testsets['POSITIVE_TIME'])
pos_recall = recall(refsets['POSITIVE_TIME'], testsets['POSITIVE_TIME'])
pos_f_measure = f_measure(refsets['POSITIVE_TIME'],
testsets['POSITIVE_TIME'])
neg_precision = precision(refsets['NEGATIVE_TIME'],
testsets['NEGATIVE_TIME'])
neg_recall = recall(refsets['NEGATIVE_TIME'], testsets['NEGATIVE_TIME'])
neg_f_measure = f_measure(refsets['NEGATIVE_TIME'],
testsets['NEGATIVE_TIME'])
print 'accuracy:', accuracy
print 'pos precision:', pos_precision
print 'pos recall:', pos_recall
print 'pos f-measure', pos_f_measure
print 'neg precision:', neg_precision
print 'neg recall:', neg_recall
print 'neg f-measure', neg_f_measure
return {
'accuracy': accuracy,
'pos precision': pos_precision,
'pos recall': pos_recall,
'pos f-measure': pos_f_measure,
'neg precision': neg_precision,
'neg recall': neg_recall,
'neg f-measure': neg_f_measure
}
def getAccuracy(self, classifier, sentences):
test_set = nltk.classify.apply_features(self.extract_features_unigram,
sentences[:500])
refsets = collections.defaultdict(set)
testsets = collections.defaultdict(set)
classifierResult = {}
for i, (feats, label) in enumerate(test_set):
refsets[label].add(i)
observed = classifier.classify(feats)
testsets[observed].add(i)
classifierResult['accuracy'] = nltk.classify.util.accuracy(
classifier, test_set)
classifierResult['suggestion precision'] = 0.0 if nltk.precision(
refsets['suggestion'],
testsets['suggestion']) == None else nltk.precision(
refsets['suggestion'], testsets['suggestion'])
classifierResult['suggestion recall'] = 0.0 if nltk.recall(
refsets['suggestion'],
testsets['suggestion']) == None else nltk.recall(
refsets['suggestion'], testsets['suggestion'])
classifierResult['nonsuggestion precision'] = 0.0 if nltk.precision(
refsets['nonsuggestion'],
testsets['nonsuggestion']) == None else nltk.precision(
refsets['nonsuggestion'], testsets['nonsuggestion'])
classifierResult['nonsuggestion recall'] = 0.0 if nltk.recall(
refsets['nonsuggestion'],
testsets['nonsuggestion']) == None else nltk.recall(
refsets['nonsuggestion'], testsets['nonsuggestion'])
return classifierResult
def evaluate_classifier(classifier):
aspects = AspectFinder.AspectFinder().get_aspects()
minus = [f for f in aspects if f[1][0]=="-"]
plus = [f for f in aspects if f[1][0]=="+"]
sentences = [' '.join(s) for s in product_reviews_1.sents()]
minusfeats = [(word_feats(s.split()), '-') for s in sentences for f in minus if s.find(f[0])!=-1]
plusfeats = [(word_feats(s.split()), '+') for s in sentences for f in plus if s.find(f[0])!=-1]
minuscutoff = int(len(minusfeats)*3/4)
pluscutoff = int(len(plusfeats)*3/4)
trainfeats = minusfeats[:minuscutoff] + plusfeats[:pluscutoff]
testfeats = minusfeats[minuscutoff:] + plusfeats[pluscutoff:]
print('train on %d instances, test on %d instances' % (len(trainfeats), len(testfeats)))
classifier = train(classifier, trainfeats)
refsets = collections.defaultdict(set)
testsets = collections.defaultdict(set)
for i, (feats, label) in enumerate(testfeats):
refsets[label].add(i)
observed = classifier.classify(feats)
testsets[observed].add(i)
print('accuracy:', nltk.classify.util.accuracy(classifier, testfeats))
print('pos precision:', precision(refsets['pos'], testsets['pos']))
print('pos recall:', recall(refsets['pos'], testsets['pos']))
print('neg precision:', precision(refsets['neg'], testsets['neg']))
print('neg recall:', recall(refsets['neg'], testsets['neg']))
classifier.show_most_informative_features()
print(classifier.classify(word_feats(["I", "hate", "it", "."])))
print(classifier.classify(word_feats(["I", "love", "it", "."])))
def precision_recall(classifier, testfeats):
""" computes precision and recall of a classifier """
refsets = collections.defaultdict(set)
testsets = collections.defaultdict(set)
for i, (feats, label) in enumerate(testfeats):
refsets[label].add(i)
observed = classifier.classify(feats)
testsets[observed].add(i)
precisions = {}
recalls = {}
for label in classifier.labels():
precisions[label] = precision(refsets[label], testsets[label])
recalls[label] = recall(refsets[label], testsets[label])
return precisions, recalls
def precision_recall(C, test_set):
"""
:param C: trained classifier
:param test_set: testing set
:return: two Dict 1st holds the precision for each label
2nd holds the recall for each label
"""
refsets = collections.defaultdict(set)
testsets = collections.defaultdict(set)
for i, (feats, label) in enumerate(test_set):
refsets[label].add(i)
observed = C.classify(feats)
testsets[observed].add(i)
precisions = {}
recalls = {}
for label in C.labels():
precisions[label] = nltk.precision(refsets[label], testsets[label])
recalls[label] = nltk.recall(refsets[label], testsets[label])
return precisions, recalls
# create training and cross-validation feature sets
trainCutoff = len(reviews) * 4/5
trainSet = reviews[:trainCutoff]
cvSet = reviews[trainCutoff:]
print ("Getting best words..")
bestwords = getBestWords(trainSet)
print ("Extracting feature sets..")
trainFeatureSet = extractFeaturesFromSet(trainSet, bestwords)
cvFeatureSet = extractFeaturesFromSet(cvSet, bestwords)
print ("Training model..")
classifier = NaiveBayesClassifier(trainFeatureSet)
refsets = collections.defaultdict(set)
testsets = collections.defaultdict(set)
for i, (feats, label) in enumerate(cvFeatureSet):
refsets[label].add(i)
observed = classifier.classify(feats)
testsets[observed].add(i)
print ("Training accuracy: ", getClassifierAccuracy(classifier, trainFeatureSet))
print ("Cross-validation accuracy: ", getClassifierAccuracy(classifier, cvFeatureSet))
print ("'pos' Precision: ", nltk.precision(refsets['pos'], testsets['pos']))
print ("'pos' Recall: ", nltk.recall(refsets['pos'], testsets['pos']))
print ("'neg' Precision: ", nltk.precision(refsets['neg'], testsets['neg']))
print ("'neg' Recall: ", nltk.recall(refsets['neg'], testsets['neg']))
classifier.showMostInformativeFeatures(20)
test_tweets = BuildFeatureVector(testing_this_round)
train_tweets = BuildFeatureVector(training_this_round)
print len(train_tweets)
print len(test_tweets)
training_set = nltk.classify.apply_features(extract_features, train_tweets)
test_set = nltk.classify.apply_features(extract_features, test_tweets)
NBClassifier = nltk.NaiveBayesClassifier.train(training_set)
#NBClassifier.show_most_informative_features(2)
TestSet(testing_this_round)
KF_metrics_accuracy.append(nltk.classify.accuracy(NBClassifier, test_set))
KF_metrics_NEU.append((precision(refSet['NEU'], testSet['NEU']), recall(refSet['NEU'], testSet['NEU']), f_measure(refSet['NEU'], testSet['NEU'])))
KF_metrics_POS.append((precision(refSet['POZ'], testSet['POZ']), recall(refSet['POZ'], testSet['POZ']), f_measure(refSet['POZ'], testSet['POZ'])))
KF_metrics_NEG.append((precision(refSet['NEG'], testSet['NEG']), recall(refSet['NEG'], testSet['NEG']), f_measure(refSet['NEG'], testSet['NEG'])))
print 'TEST accuracy:', sum(KF_metrics_accuracy) / float(len(KF_metrics_accuracy))
print ''
print ''
print 'NEU precision:', sum(KF_metrics_NEU[0]) / float(len(KF_metrics_NEU[0]))
print 'NEU recall:', sum(KF_metrics_NEU[1]) / float(len(KF_metrics_NEU[1]))
print 'NEU F-measure:', sum(KF_metrics_NEU[2]) / float(len(KF_metrics_NEU[2]))
print ''
print 'POS precision:', sum(KF_metrics_POS[0]) / float(len(KF_metrics_POS[0]))
print 'POS recall:', sum(KF_metrics_POS[1]) / float(len(KF_metrics_POS[1]))
print 'POS F-measure:', sum(KF_metrics_POS[2]) / float(len(KF_metrics_POS[2]))
print ''
print 'NEG precision:', sum(KF_metrics_NEG[0]) / float(len(KF_metrics_NEG[0]))
def evaluate_features(feature_select):
posFeatures_train = []
negFeatures_train = []
posFeatures_test = []
negFeatures_test = []
#http://stackoverflow.com/questions/367155/splitting-a-string-into-words-and-punctuation
#breaks up the sentences into lists of individual words (as selected by the input mechanism) and appends 'pos' or 'neg' after each list
with open(POS_FILE_TRAIN, "rb") as f:
posSentences_train = f.readlines()
#posSentences = pos_data.split('\n')
random.shuffle(posSentences_train)
with open(POS_FILE_TEST, "rb") as f:
posSentences_test = f.readlines()
#posSentences = pos_data.split('\n')
random.shuffle(posSentences_test)
with open(NEG_FILE_TRAIN, "rb") as f:
#negSentences = f.read().split('\n')
negSentences_train = f.readlines()
random.shuffle(negSentences_train)
with open(NEG_FILE_TEST, "rb") as f:
#negSentences = f.read().split('\n')
negSentences_test = f.readlines()
random.shuffle(negSentences_test)
#with open(RT_POLARITY_POS_FILE, 'r') as posSentences:
for i in posSentences_train:
posWords = re.findall(r"[\w']+|[.,!?;@#]", str(i).rstrip())
posWords = [feature_select(posWords), 'pos'] #pos = contains location
posFeatures_train.append(posWords)
str_i = (i.decode("utf-8")).strip()
Train_twit_Dic[frozenset(posWords[0].items())] = str_i
for i in posSentences_test:
posWords_test = re.findall(r"[\w']+|[.,!?;@#]", str(i).rstrip())
posWords_test = [feature_select(posWords_test),
'pos'] #pos = contains location
posFeatures_test.append(posWords_test)
str1 = (i.decode("utf-8")).strip()
Test_twit_Dic[frozenset(posWords_test[0].items())] = str1
#with open(RT_POLARITY_NEG_FILE, 'r') as negSenBufferedReader: <_io.BufferedReader name='/home/ira/Dropbox/twitter/contain_location_tweets.txt'>tences:
for i in negSentences_train:
negWords = re.findall(r"[\w']+|[.,!?;@#]", str(i).rstrip())
negWords = [feature_select(negWords),
'neg'] #neg = doesn't contain location
negFeatures_train.append(negWords)
str2 = (i.decode("utf-8")).strip()
Train_twit_Dic[frozenset(negWords[0].items())] = str2
for i in negSentences_test:
negWords_test = re.findall(r"[\w']+|[.,!?;@#]", str(i).rstrip())
negWords_test = [feature_select(negWords_test),
'neg'] #neg = doesn't contain location
negFeatures_test.append(negWords_test)
str3 = (i.decode("utf-8")).strip()
Test_twit_Dic[frozenset(negWords_test[0].items())] = str3
#selects 3/4 of the features to be used for training and 1/4 to be used for testing
#posCutoff = int(math.floor(len(posFeatures)*3/4))
#negCutoff = int(math.floor(len(negFeatures)*3/4))
#trainFeatures = posFeatures[:posCutoff] + negFeatures[:negCutoff] ###need to understand what is test here
#testFeatures = posFeatures[posCutoff:] + negFeatures[negCutoff:]
trainFeatures = posFeatures_train + negFeatures_train ###need to understand what is test here
testFeatures = posFeatures_test + negFeatures_test
##############################################################################3
#trains a Naive Bayes Classifier
classifier = NaiveBayesClassifier.train(trainFeatures)
#initiates referenceSets and testSets
referenceSets = collections.defaultdict(set)
testSets = collections.defaultdict(set)
#puts correctly labeled sentences in referenceSets and the predictively labeled version in testsets
for i, (features, label) in enumerate(testFeatures):
referenceSets[label].add(i)
predicted = classifier.classify(features)
testSets[predicted].add(i)
#################### MINE ####################################
if predicted == "pos": ##the twit according to the classifier contains a location
twiit = Test_twit_Dic[frozenset(features.items())]
list_close_twits = Close_Twt_Dic[twiit]
words = nltk.word_tokenize(twiit)
tagged_words = ner_tagger.tag(words)
lbl = ""
for tag_w in tagged_words:
if tag_w[1] == "LOCATION":
lbl = lbl + tag_w[0] + " " #found a label for our twiit
final_lbl = lbl
### employ satnford trained classifier on all of the close twiits to find
if lbl == "": #couldnt find the location (lable) for our twiit, lets try to find it wihitn its physical neiborhood twwits
lbl_list = []
for s in list_close_twits:
words = nltk.word_tokenize(s)
tagged_words = ner_tagger.tag(words)
lbl = ""
for tag_w in tagged_words:
if tag_w[1] == "LOCATION":
lbl = lbl + tag_w[0] + " "
if lbl != "":
lbl_list.append(lbl)
## find most common str (label) in lbl_list
c = Counter(lbl_list)
c.most_common(1)
final_lbl = c[0][0]
#prints metrics to show how well the feature selection did
print('train on %d instances, test on %d instances' %
(len(trainFeatures), len(testFeatures)))
print('accuracy:', nltk.classify.util.accuracy(classifier, testFeatures))
#print 'pos precision:', nltk.metrics.precision(referenceSets['pos'], testSets['pos'])
print('pos precision:', precision(referenceSets['pos'], testSets['pos']))
print('pos recall:', recall(referenceSets['pos'], testSets['pos']))
#print 'neg precision:', nltk.metrics.precision(referenceSets['neg'], testSets['neg'])
print('neg precision:', precision(referenceSets['neg'], testSets['neg']))
print('neg recall:', recall(referenceSets['neg'], testSets['neg']))
classifier.show_most_informative_features(10)
def evaluate_classifier(featx):
negfeats = [(featx(f), 'neg') for f in word_split(negdata)]
posfeats = [(featx(f), 'pos') for f in word_split(posdata)]
negcutoff = len(negfeats)*3/4
poscutoff = len(posfeats)*3/4
trainfeats = negfeats[:negcutoff] + posfeats[:poscutoff]
testfeats = negfeats[negcutoff:] + posfeats[poscutoff:]
# using 3 classifiers
classifier_list = ['nb', 'svm']
#classifier_list = ['nb', 'maxent', 'svm']
for cl in classifier_list:
#if cl == 'maxent':
# classifierName = 'Maximum Entropy'
# classifier = MaxentClassifier.train(trainfeats, 'GIS', trace=0, encoding=None, labels=None, sparse=True, gaussian_prior_sigma=0, max_iter = 1)
if cl == 'svm':
classifierName = 'SVM'
classifier = SklearnClassifier(LinearSVC(), sparse=False)
classifier.train(trainfeats)
else:
classifierName = 'Naive Bayes'
classifier = NaiveBayesClassifier.train(trainfeats)
refsets = collections.defaultdict(set)
testsets = collections.defaultdict(set)
for i, (feats, label) in enumerate(testfeats):
refsets[label].add(i)
observed = classifier.classify(feats)
testsets[observed].add(i)
accuracy = nltk.classify.util.accuracy(classifier, testfeats)
pos_precision = precision(refsets['pos'], testsets['pos'])
pos_recall = recall(refsets['pos'], testsets['pos'])
pos_fmeasure = f_measure(refsets['pos'], testsets['pos'])
neg_precision = precision(refsets['neg'], testsets['neg'])
neg_recall = recall(refsets['neg'], testsets['neg'])
neg_fmeasure = f_measure(refsets['neg'], testsets['neg'])
print ''
print '---------------------------------------'
print 'SINGLE FOLD RESULT ' + '(' + classifierName + ')'
print '---------------------------------------'
print 'accuracy:', accuracy
print 'precision', (pos_precision + neg_precision) / 2
print 'recall', (pos_recall + neg_recall) / 2
print 'f-measure', (pos_fmeasure + neg_fmeasure) / 2
#classifier.show_most_informative_features()
print ''
## CROSS VALIDATION
trainfeats = negfeats + posfeats
# SHUFFLE TRAIN SET
# As in cross validation, the test chunk might have only negative or only positive data
random.shuffle(trainfeats)
n = 5 # 5-fold cross-validation
for cl in classifier_list:
subset_size = len(trainfeats) / n
accuracy = []
pos_precision = []
pos_recall = []
neg_precision = []
neg_recall = []
pos_fmeasure = []
neg_fmeasure = []
cv_count = 1
for i in range(n):
testing_this_round = trainfeats[i*subset_size:][:subset_size]
training_this_round = trainfeats[:i*subset_size] + trainfeats[(i+1)*subset_size:]
#if cl == 'maxent':
# classifierName = 'Maximum Entropy'
# classifier = MaxentClassifier.train(training_this_round, 'GIS', trace=0, encoding=None, labels=None, sparse=True, gaussian_prior_sigma=0, max_iter = 1)
if cl == 'svm':
classifierName = 'SVM'
classifier = SklearnClassifier(LinearSVC(), sparse=False)
classifier.train(training_this_round)
else:
classifierName = 'Naive Bayes'
classifier = NaiveBayesClassifier.train(training_this_round)
refsets = collections.defaultdict(set)
testsets = collections.defaultdict(set)
for i, (feats, label) in enumerate(testing_this_round):
refsets[label].add(i)
observed = classifier.classify(feats)
testsets[observed].add(i)
cv_accuracy = nltk.classify.util.accuracy(classifier, testing_this_round)
cv_pos_precision = precision(refsets['pos'], testsets['pos'])
cv_pos_recall = recall(refsets['pos'], testsets['pos'])
cv_pos_fmeasure = f_measure(refsets['pos'], testsets['pos'])
cv_neg_precision = precision(refsets['neg'], testsets['neg'])
cv_neg_recall = recall(refsets['neg'], testsets['neg'])
cv_neg_fmeasure = f_measure(refsets['neg'], testsets['neg'])
accuracy.append(cv_accuracy)
pos_precision.append(cv_pos_precision)
pos_recall.append(cv_pos_recall)
neg_precision.append(cv_neg_precision)
neg_recall.append(cv_neg_recall)
pos_fmeasure.append(cv_pos_fmeasure)
neg_fmeasure.append(cv_neg_fmeasure)
cv_count += 1
print '---------------------------------------'
print 'N-FOLD CROSS VALIDATION RESULT ' + '(' + classifierName + ')'
print '---------------------------------------'
print 'accuracy:', sum(accuracy) / n
print 'precision', (sum(pos_precision)/n + sum(neg_precision)/n) / 2
print 'recall', (sum(pos_recall)/n + sum(neg_recall)/n) / 2
print 'f-measure', (sum(pos_fmeasure)/n + sum(neg_fmeasure)/n) / 2
print ''
testSets['pos'] = set()
testSets['neg'] = set()
shuffle(test)
for i, (features, label) in enumerate(test):
referenceSets[label].add(i)
predicted = classifier.classify(features)
testSets[predicted].add(i)
print 'After training on %d samples, start to test on %d instances:' % (
len(train), len(test))
print 'accuracy: %.2f' % nltk.classify.util.accuracy(classifier, test)
print 'pos precision: %.2f' % nltk.precision(referenceSets['pos'],
testSets['pos'])
print 'neg precision: %.2f' % nltk.precision(referenceSets['neg'],
testSets['neg'])
print 'pos recall: %.2f' % nltk.recall(referenceSets['pos'],
testSets['pos'])
print 'neg recall: %.2f' % nltk.recall(referenceSets['neg'],
testSets['neg'])
testSets['neg'].clear()
testSets['pos'].clear()
referenceSets['neg'].clear()
referenceSets['pos'].clear()
for j in range(5):
test_subset = test[j * 100:(j + 1) * 100]
for i, (features, label) in enumerate(test_subset):
referenceSets[label].add(i)
predicted = classifier.classify(features)
testSets[predicted].add(i)
minus3feats[:minus3cutoff] + plus3feats[:plus3cutoff]
testfeats = minus1feats[minus1cutoff:] + plus1feats[plus1cutoff:] + \
minus2feats[minus2cutoff:] + plus2feats[plus2cutoff:] + \
minus3feats[minus3cutoff:] + plus3feats[plus3cutoff:]
print('train on %d instances, test on %d instances' % (len(trainfeats), len(testfeats)))
classifier = NaiveBayesClassifier.train(trainfeats)
refsets = collections.defaultdict(set)
testsets = collections.defaultdict(set)
for i, (feats, label) in enumerate(testfeats):
refsets[label].add(i)
observed = classifier.classify(feats)
testsets[observed].add(i)
print('accuracy:', nltk.classify.util.accuracy(classifier, testfeats))
print('+1 precision:', precision(refsets['+1'], testsets['+1']))
print('+1 recall:', recall(refsets['+1'], testsets['+1']))
print('+2 precision:', precision(refsets['+2'], testsets['+2']))
print('+2 recall:', recall(refsets['+2'], testsets['+2']))
print('+3 precision:', precision(refsets['+3'], testsets['+3']))
print('+3 recall:', recall(refsets['+3'], testsets['+3']))
print('-1 precision:', precision(refsets['-1'], testsets['-1']))
print('-1 recall:', recall(refsets['-1'], testsets['-1']))
print('-2 precision:', precision(refsets['-2'], testsets['-2']))
print('-2 recall:', recall(refsets['-2'], testsets['-2']))
print('-3 precision:', precision(refsets['-3'], testsets['-3']))
print('-3 recall:', recall(refsets['-3'], testsets['-3']))
classifier.show_most_informative_features()
print(classifier.classify(word_feats(["I", "hate", "it", "."])))
print(classifier.classify(word_feats(["I", "love", "it", "."])))
def evaluate_classifier(featx):
negfeats = [(featx(f), 'neg') for f in word_split(negdata)]
posfeats = [(featx(f), 'pos') for f in word_split(posdata)]
#print(negfeats)
negcutoff = int(len(negfeats) * 3 / 4)
poscutoff = int(len(posfeats) * 3 / 4)
#print(negcutoff)
trainfeats = negfeats[:negcutoff] + posfeats[:poscutoff]
#print(trainfeats)
testfeats = negfeats[negcutoff:] + posfeats[poscutoff:]
classifierName = 'SVM'
classifier = SklearnClassifier(LinearSVC(), sparse=False).train(trainfeats)
#classifier.train(trainfeats)
refsets = collections.defaultdict(set)
testsets = collections.defaultdict(set)
#print(testfeats)
for i, (feats, label) in enumerate(testfeats):
#feats : list of words
#label : neg/pos
#observed : neg/pos
#print(feats,'---',label)
refsets[label].add(i)
observed = classifier.classify(feats)
#print(observed)
testsets[observed].add(i)
accuracy = nltk.classify.util.accuracy(classifier, testfeats)
pos_precision = nltk.precision(refsets['pos'], testsets['pos'])
pos_recall = nltk.recall(refsets['pos'], testsets['pos'])
pos_fmeasure = nltk.f_measure(refsets['pos'], testsets['pos'])
neg_precision = nltk.precision(refsets['neg'], testsets['neg'])
neg_recall = nltk.recall(refsets['neg'], testsets['neg'])
neg_fmeasure = nltk.f_measure(refsets['neg'], testsets['neg'])
print('')
print('---------------------------------------')
print('SINGLE FOLD RESULT ' + '(' + classifierName + ')')
print('---------------------------------------')
print('accuracy:', accuracy)
print('precision', (pos_precision + neg_precision) / 2)
print('recall', (pos_recall + neg_recall) / 2)
print('f-measure', (pos_fmeasure + neg_fmeasure) / 2)
#classifier.show_most_informative_features()
print('')
## CROSS VALIDATION
trainfeats = negfeats + posfeats
# SHUFFLE TRAIN SET
# As in cross validation, the test chunk might have only negative or only positive data
random.shuffle(trainfeats)
n = 5 # 5-fold cross-validation
subset_size = int(len(trainfeats) / n)
accuracy = []
pos_precision = []
pos_recall = []
neg_precision = []
neg_recall = []
pos_fmeasure = []
neg_fmeasure = []
cv_count = 1
for i in range(n):
testing_this_round = trainfeats[i * subset_size:][:subset_size]
training_this_round = trainfeats[:i * subset_size] + trainfeats[
(i + 1) * subset_size:]
classifierName = 'SVM'
classifier = SklearnClassifier(LinearSVC(), sparse=False)
classifier.train(training_this_round)
refsets = collections.defaultdict(set)
testsets = collections.defaultdict(set)
for i, (feats, label) in enumerate(testing_this_round):
refsets[label].add(i)
observed = classifier.classify(feats)
testsets[observed].add(i)
cv_accuracy = nltk.classify.util.accuracy(classifier,
testing_this_round)
cv_pos_precision = nltk.precision(refsets['pos'], testsets['pos'])
cv_pos_recall = nltk.recall(refsets['pos'], testsets['pos'])
cv_pos_fmeasure = nltk.f_measure(refsets['pos'], testsets['pos'])
cv_neg_precision = nltk.precision(refsets['neg'], testsets['neg'])
cv_neg_recall = nltk.recall(refsets['neg'], testsets['neg'])
cv_neg_fmeasure = nltk.f_measure(refsets['neg'], testsets['neg'])
accuracy.append(cv_accuracy)
pos_precision.append(cv_pos_precision)
pos_recall.append(cv_pos_recall)
neg_precision.append(cv_neg_precision)
neg_recall.append(cv_neg_recall)
pos_fmeasure.append(cv_pos_fmeasure)
neg_fmeasure.append(cv_neg_fmeasure)
cv_count += 1
print('---------------------------------------')
print('N-FOLD CROSS VALIDATION RESULT ' + '(' + classifierName + ')')
print('---------------------------------------')
print('accuracy:', sum(accuracy) / n)
print('precision', (sum(pos_precision) / n + sum(neg_precision) / n) / 2)
print('recall', (sum(pos_recall) / n + sum(neg_recall) / n) / 2)
print('f-measure', (sum(pos_fmeasure) / n + sum(neg_fmeasure) / n) / 2)
print('')
minusfeats.append(feat)
if len(minusfeats) > len(plusfeats):
minusfeats = minusfeats[:len(plusfeats)]
else:
plusfeats = plusfeats[:len(minusfeats)]
minuscutoff = int(len(minusfeats)*3/4)
pluscutoff = int(len(plusfeats)*3/4)
trainfeats = minusfeats[:minuscutoff] + plusfeats[:pluscutoff]
testfeats = minusfeats[minuscutoff:] + plusfeats[pluscutoff:]
print('train on %d instances, test on %d instances' % (len(trainfeats), len(testfeats)))
classifier = NaiveBayesClassifier.train(trainfeats)
refsets = collections.defaultdict(set)
testsets = collections.defaultdict(set)
for i, (feats, label) in enumerate(testfeats):
refsets[label].add(i)
observed = classifier.classify(feats)
testsets[observed].add(i)
print('accuracy:', nltk.classify.util.accuracy(classifier, testfeats))
print('pos precision:', precision(refsets['+'], testsets['+']))
print('pos recall:', recall(refsets['+'], testsets['+']))
print('neg precision:', precision(refsets['-'], testsets['-']))
print('neg recall:', recall(refsets['-'], testsets['-']))
classifier.show_most_informative_features()
print(classifier.classify(word_feats(["I", "hate", "it", "."])))
print(classifier.classify(word_feats(["I", "love", "it", "."])))
# Training using the NaiveBayes algorithm
train_set, test_set = train_test_split(featuresets,
test_size=0.33,
random_state=42)
base_classifier = nltk.NaiveBayesClassifier.train(train_set)
print('Accuracy:', nltk.classify.accuracy(base_classifier, test_set))
# Evaluation
refsets = collections.defaultdict(set)
testsets = collections.defaultdict(set)
for i, (feats, label) in enumerate(test_set):
refsets[label].add(i)
observed = base_classifier.classify(feats)
testsets[observed].add(i)
print('Precision:', nltk.precision(refsets['pos'], testsets['pos']))
print('Recall:', nltk.recall(refsets['pos'], testsets['pos']))
print('f_measure:', nltk.f_measure(refsets['pos'], testsets['pos']))
# In[6]:
## Using the lexicon file (Subjectivity) and define the feature
def readSubjectivity(path):
flexicon = open(path, 'r')
# initialize an empty dictionary
sldict = {}
for line in flexicon:
fields = line.split() # default is to split on whitespace
# split each field on the '=' and keep the second part as the value
strength = fields[0].split("=")[1]
word = fields[2].split("=")[1]
NBayesClassifier=nltk.NaiveBayesClassifier.train(trainFeat)
#NBResultLabels=[NBayesClassifier.classify(extract_features(tweet[0])) for tweet in testData]
print("Accuracy : " + str(nltk.classify.util.accuracy(NBayesClassifier, testFeat)*100) + " %")
refsets = collections.defaultdict(set)
testsets = collections.defaultdict(set)
for i, (feats, label) in enumerate(testFeat):
refsets[label].add(i)
observed = NBayesClassifier.classify(feats)
testsets[observed].add(i)
print 'sarcasm precision : ' + str((precision(refsets['sarcasm'], testsets['sarcasm'])*100)) + " %"
print 'sarcasm recall : ' + str((recall(refsets['sarcasm'], testsets['sarcasm'])*100)) + " %"
print 'sarcasm F-measure : ' + str((f_measure(refsets['sarcasm'], testsets['sarcasm'])*100)) + " %"
print 'non-sarcasm precision : ' + str((precision(refsets['non-sarcasm'], testsets['non-sarcasm'])*100)) + " %"
print 'non-sarcasm recall : ' + str((recall(refsets['non-sarcasm'], testsets['non-sarcasm'])*100)) + " %"
print 'non-sarcasm F-measure : ' + str((f_measure(refsets['non-sarcasm'], testsets['non-sarcasm'])*100)) + " %"
#NBayesClassifier.show_most_informative_features(100)
# print(NBResultLabels)
# if NBResultLabels.count('positive')>NBResultLabels.count('negative'):
# print "NB Result Sarcastic Sentiment\t\t:" + str(100*NBResultLabels.count('sarcasm')/len(NBResultLabels))+"%"
# else:
# print "NB Result Non-Sarcastic Sentiment\t:" + str(100*NBResultLabels.count('non-sarcasm')/len(NBResultLabels))+"%"
