Source: View original notebook on GitHub
Category: Machine Learning / Learn ML
Naive Bayes for Text Classification (Multinomial Naive Bayes)
# lets get some sample text using NLTK -> watch out NlP notebook for detailed NLP preocesses
from nltk.tokenize import word_tokenize
from nltk.corpus import stopwords
from nltk.stem import SnowballStemmer
import string
def NLP_preprocess(text):
text = text.lower()
# step 1 - wordTokenize
li = word_tokenize(text)
# step 2 - stopword removal
stopword_list = stopwords.words('english')
li = [word for word in li if word not in stopword_list]
# step 3 - puntuation Removal
punch_list = string.punctuation
li = [word for word in li if word not in punch_list]
# step 4 - Lemmatiziation
ss = SnowballStemmer('english')
li = [ss.stem(word) for word in li ]
return li
Let's see Naive Bayes in action predicting whether the review is positive or negative
text = ['Best movie ever made by Marvel. Loved every detail of it. ',
'It could not get better ,satisfied Ending as well. May god Bless Tony.',
'movie was not upto the mark ,I am disappointed . why tony stark have to die in the end. why Bad Things happens to good people',
"I have never seen so many people get up to pee in one film and I could not believe they were sacrificing even a second, because Endgame is incredible.",
"OK. It's great. Obviously. Emotional, funny, epic, all those good things. Five stars, eight thumbs up.",
"There is so much wrong with Endgame that I can not get out of my head. Primarily, time travel removes all jeopardy from all future stories.",
]
Y = [1,1,0,1,1,0] # 1 is good review , 0 is bad review
print(len(text))
Output:
6
# Preprocesing- not necessary ; could have done using CountVectorizer() by passing it to tokenizer keyworded argument
# X = []
# for i in text :
# X.append(' '.join(NLP_preprocess(i)))
# print(X)
# vectorization
from sklearn.feature_extraction.text import CountVectorizer
cv = CountVectorizer(tokenizer=NLP_preprocess,ngram_range=(1,1))
# cv.fit(X)
# cv.transform(X).toarray()
# there are two function fit and transform which are combined in fit_transform
arr = cv.fit_transform(text).toarray()
arr
Output:
array([[0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 1, 0, 1, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 1, 1, 1, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 0],
[0, 1, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 1, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1,
0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 0, 1, 0, 1, 0, 0],
[0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 0, 1, 0, 0, 0,
1, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 1, 0, 0, 1, 1, 1,
0, 0, 1, 0, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 1, 1, 0,
0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0,
0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1,
1, 0, 0, 0, 0, 1, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0,
1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 0, 1, 0, 0, 1]], dtype=int64)
print(cv.vocabulary_)
Output:
{'best': 3, 'movi': 36, 'ever': 16, 'made': 31, 'marvel': 34, 'love': 30, 'everi': 17, 'detail': 7, 'could': 6, 'get': 22, 'better': 4, 'satisfi': 47, 'end': 12, 'well': 59, 'may': 35, 'god': 23, 'bless': 5, 'toni': 56, 'upto': 58, 'mark': 33, 'disappoint': 9, 'stark': 51, 'die': 8, 'bad': 1, 'thing': 53, 'happen': 26, 'good': 24, 'peopl': 43, 'never': 38, 'seen': 49, 'mani': 32, 'pee': 42, 'one': 41, 'film': 18, 'believ': 2, 'sacrif': 46, 'even': 15, 'second': 48, 'endgam': 13, 'incred': 28, 'ok.': 40, "'s": 0, 'great': 25, 'obvious': 39, 'emot': 11, 'funni': 20, 'epic': 14, 'five': 19, 'star': 50, 'eight': 10, 'thumb': 54, 'much': 37, 'wrong': 60, 'head': 27, 'primarili': 44, 'time': 55, 'travel': 57, 'remov': 45, 'jeopardi': 29, 'futur': 21, 'stori': 52}
cv.vocabulary_['bad']
Output:
1
# lets verify by inverse _transformation
text[0]
Output:
'Best movie ever made by Marvel. Loved every detail of it. '
cv.inverse_transform(arr[0]) # same thing
Output:
[array(['best', 'detail', 'ever', 'everi', 'love', 'made', 'marvel',
'movi'], dtype='<U10')]
# not doing TfidfTransformer here see Assignment 4 ,applies there
Predicting for example: ' It is difficult for me to put into words how perfectly realized Avengers: Endgame is. '
X = arr
X.shape
Output:
(6, 61)
Y = Y
Y
Output:
[1, 1, 0, 1, 1, 0]
from sklearn.naive_bayes import MultinomialNB
mnb = MultinomialNB()
mnb.fit(X,Y)
Output:
MultinomialNB(alpha=1.0, class_prior=None, fit_prior=True)
test = ['It is difficult for me to put into words how perfectly realized Avengers: Endgame is. ',
'Bad movie overall.',
'Worst movie ever', # will be wrongly predicted because it is not seen in training data
'Jaw-Dropping movie ',# (correct by luck maybe)
'Good movie overall.']
arr = cv.transform(test)
arr.toarray()
Output:
array([[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]], dtype=int64)
mnb.predict(arr) # yes test[0] was bas review and test[1] was good
Output:
array([1, 0, 1, 1, 1])