import nltk
from nltk import word_tokenize
from nltk import WordNetLemmatizer
from collections import Counter
from nltk import NaiveBayesClassifier, classify
def Process(data):
lemmatizer = WordNetLemmatizer()
return [lemmatizer.lemmatize(word.lower()) for word in word_tokenize(unicode(sentence,errors='ignore'))]
def Features_Extraction(text, setting):
if setting=='bow':
# Bow means bag-of-words
return {word: count for word, count in Counter(Process(text)).items() if not word in stop}
else:
return {word: True for word in Process(text) if not word in stop}
features = [(Features_Extraction(email, 'bow'), label) for (email, label)
in emails]
def training_Model (Features, samples):
Size = int(len(Features) * samples)
training , testing = Features[:Size], Features[Size:]
print ('Training = ' + str(len(training)) + ' emails')
print ('Testing = ' + str(len(testing)) + ' emails')
classifier = NaiveBayesClassifier.train(training)
def evaluate(training, tesing, classifier):
print ('Training Accuracy is ' + str(classify.accuracy(classifier,train_set)))
print ('Testing Accuracy i ' + str(classify.accuracy(classifier,test_set)))
