Machine Learning using Python

Chapter 10: Text Analytics

10.2 Sentiment Classification

10.2.1 Loading the dataset

import pandas as pd
import numpy as np

import warnings
[Link]('ignore')

train_ds = pd.read_csv( "sentiment_train", delimiter="\t" )

train_ds.head( 5 )

sentiment text

0 1 The Da Vinci Code book is just awesome.

1 1 this was the first clive cussler i've ever rea...

2 1 i liked the Da Vinci Code a lot.

3 1 i liked the Da Vinci Code a lot.

4 1 I liked the Da Vinci Code but it ultimatly did...

pd.set_option('max_colwidth', 800)
train_ds[train_ds.sentiment == 1][0:5]

sentiment text

0 1 The Da Vinci Code book is just awesome.

this was the first clive cussler i've ever read, but even books like Relic, and
1 1
Da Vinci code were more plausible than this.

2 1 i liked the Da Vinci Code a lot.

3 1 i liked the Da Vinci Code a lot.

4 1 I liked the Da Vinci Code but it ultimatly didn't seem to hold it's own.

Copyright © 2019 by Wiley India Pvt. Ltd. 1/13

 Machine Learning using Python
train_ds[train_ds.sentiment == 0][0:5]

sentiment text

3943 0 da vinci code was a terrible movie.

Then again, the Da Vinci code is super shitty movie, and it made like
3944 0
700 million.

3945 0 The Da Vinci Code comes out tomorrow, which sucks.

3946 0 i thought the da vinci code movie was really boring.

God, Yahoo Games has this truly-awful looking Da Vinci Code-themed

3947 0
skin on it's chessboard right now.

10.2.2 Exploring the dataset

train_ds.info()

<class '[Link]'>
RangeIndex: 6918 entries, 0 to 6917
Data columns (total 2 columns):
sentiment 6918 non-null int64
text 6918 non-null object
dtypes: int64(1), object(1)
memory usage: 108.2+ KB

Copyright © 2019 by Wiley India Pvt. Ltd. 2/13

 Machine Learning using Python
import [Link] as plt
import seaborn as sn
%matplotlib inline

[Link]( figsize=(6,5))
# create count plot
ax = [Link](x='sentiment', data=train_ds)
# annotate
for p in [Link]:
[Link](p.get_height(), (p.get_x()+0.1, p.get_height()+50))

10.2.3 Text Preprocessing

[Link] Creating Count Vectors for sentiment_train dataset

Copyright © 2019 by Wiley India Pvt. Ltd. 3/13

 Machine Learning using Python
from sklearn.feature_extraction.text import CountVectorizer

# Initialize the CountVectorizer

count_vectorizer = CountVectorizer()
# Create the dictionary from the corpus
feature_vector = count_vectorizer.fit( train_ds.text )
# Get the feature names
features = feature_vector.get_feature_names()
print( "Total number of features: ", len(features))

Total number of features: 2132

import random

[Link](features, 10)

['mad',
'ew',
'sceneries',
'awesome',
'gary',
'wept',
'hope',
'life',
'television',
'aimee']

train_ds_features = count_vectorizer.transform( train_ds.text )

type(train_ds_features)

[Link].csr_matrix

train_ds_features.shape

(6918, 2132)

train_ds_features.getnnz()

65398

print( "Density of the matrix: ",

train_ds_features.getnnz() * 100 /
(train_ds_features.shape[0] * train_ds_features.shape[1]))

Density of the matrix: 0.4434010415225908

[Link] Displaying Document Vectors

# Converting the matrix to a dataframe

train_ds_df = [Link](train_ds_features.todense())
# Setting the column names to the features i.e. words
train_ds_df.columns = features

Copyright © 2019 by Wiley India Pvt. Ltd. 4/13

 Machine Learning using Python
train_ds[0:1]

sentiment text

0 1 The Da Vinci Code book is just awesome.

train_ds_df.iloc[0:1, 150:157]

away awesome awesomely awesomeness awesomest awful awkward

0 0 1 0 0 0 0 0

train_ds_df[['the', 'da', "vinci", "code", "book", 'is', 'just', 'awesome']][0:1

]

the da vinci code book is just awesome

0 1 1 1 1 1 1 1 1

[Link] Removing low frequency words

# summing up the occurances of features column wise

features_counts = [Link]( train_ds_features.toarray(), axis = 0 )
feature_counts_df = [Link]( dict( features = features,
counts = features_counts ) )

[Link]( figsize=(12,5))
[Link](feature_counts_df.counts, bins=50, range = (0, 2000));
[Link]( 'Frequency of words' )
[Link]( 'Density' );

len(feature_counts_df[feature_counts_df.counts == 1])

1228

Copyright © 2019 by Wiley India Pvt. Ltd. 5/13

 Machine Learning using Python
# Initialize the CountVectorizer
count_vectorizer = CountVectorizer(max_features=1000)
# Create the dictionary from the corpus
feature_vector = count_vectorizer.fit( train_ds.text )
# Get the feature names
features = feature_vector.get_feature_names()
# Transform the document into vectors
train_ds_features = count_vectorizer.transform( train_ds.text )
# Count the frequency of the features
features_counts = [Link]( train_ds_features.toarray(), axis = 0 )
feature_counts = [Link]( dict( features = features,
counts = features_counts ) )

feature_counts.sort_values('counts',
ascending = False)[0:15]

counts features

866 3306 the

37 2154 and

358 2093 harry

675 2093 potter

138 2002 code

934 2001 vinci

178 2001 da

528 2000 mountain

104 2000 brokeback

488 1624 love

423 1520 is

941 1176 was

60 1127 awesome

521 1094 mission

413 1093 impossible

[Link] Removing Stop Words

from sklearn.feature_extraction import text

my_stop_words = text.ENGLISH_STOP_WORDS

#Printing first few stop words

print("Few stop words: ", list(my_stop_words)[0:10])

Few stop words: ['twenty', 'after', 'i', 'first', 'un', 'her', 'the
reupon', 'meanwhile', 'then', 'moreover']

Copyright © 2019 by Wiley India Pvt. Ltd. 6/13

 Machine Learning using Python
# Adding custom words to the list of stop words
my_stop_words = text.ENGLISH_STOP_WORDS.union( ['harry', 'potter', 'code', 'vinc
i', 'da',
'harry', 'mountain', 'movie', 'm
ovies'])

[Link] Creating Count Vectors

# Setting stop words list

count_vectorizer = CountVectorizer( stop_words = my_stop_words,
max_features = 1000 )
feature_vector = count_vectorizer.fit( train_ds.text )
train_ds_features = count_vectorizer.transform( train_ds.text )
features = feature_vector.get_feature_names()
features_counts = [Link]( train_ds_features.toarray(), axis = 0 )
feature_counts = [Link]( dict( features = features,
counts = features_counts ) )

feature_counts.sort_values( "counts", ascending = False )[0:15]

counts features

73 2000 brokeback

408 1624 love

39 1127 awesome

436 1094 mission

341 1093 impossible

390 974 like

745 602 sucks

743 600 sucked

297 578 hate

652 374 really

741 365 stupid

362 287 just

374 276 know

742 276 suck

409 256 loved

Copyright © 2019 by Wiley India Pvt. Ltd. 7/13

 Machine Learning using Python
from [Link] import PorterStemmer

stemmer = PorterStemmer()
analyzer = CountVectorizer().build_analyzer()

#Custom function for stemming and stop word removal

def stemmed_words(doc):
### Stemming of words
stemmed_words = ([Link](w) for w in analyzer(doc))
### Remove the words in stop words list
non_stop_words = [ word for word in list(set(stemmed_words) - set(my_stop_wo
rds)) ]
return non_stop_words

count_vectorizer = CountVectorizer( analyzer=stemmed_words,

max_features = 1000)
feature_vector = count_vectorizer.fit( train_ds.text )
train_ds_features = count_vectorizer.transform( train_ds.text )
features = feature_vector.get_feature_names()
features_counts = [Link]( train_ds_features.toarray(), axis = 0 )
feature_counts = [Link]( dict( features = features,
counts = features_counts ) )
feature_counts.sort_values( "counts", ascending = False )[0:15]

counts features

80 1930 brokeback

297 1916 harri

407 1837 love

803 1378 suck

922 1142 wa

43 1116 awesom

345 1090 imposs

433 1090 mission

439 1052 movi

393 823 like

299 636 hate

54 524 becaus

604 370 realli

796 364 stupid

379 354 know

[Link] Distribution of words across different sentiment

Copyright © 2019 by Wiley India Pvt. Ltd. 8/13

 Machine Learning using Python
# Convert the document vector matrix into dataframe
train_ds_df = [Link](train_ds_features.todense())
# Assign the features names to the column
train_ds_df.columns = features
# Assign the sentiment labels to the train_ds
train_ds_df['sentiment'] = train_ds.sentiment

[Link]( x = 'sentiment', y = 'awesom', data = train_ds_df, estimator=sum );

[Link]( x = 'sentiment', y = 'realli', data = train_ds_df, estimator=sum );

Copyright © 2019 by Wiley India Pvt. Ltd. 9/13

 Machine Learning using Python
[Link]( x = 'sentiment', y = 'hate', data = train_ds_df, estimator=sum );

10.3 Naive Bayes Model for Sentiment Classification

10.3.1 Split the dataset

from sklearn.model_selection import train_test_split

train_X, test_X, train_y, test_y = train_test_split( train_ds_features,

train_ds.sentiment,
test_size = 0.3,
random_state = 42 )

10.3.2 Build Naive Bayes Model

from sklearn.naive_bayes import BernoulliNB

nb_clf = BernoulliNB()
nb_clf.fit( train_X.toarray(), train_y )

BernoulliNB(alpha=1.0, binarize=0.0, class_prior=None, fit_prior=Tru

e)

10.3.3 Make prediction on test case

Copyright © 2019 by Wiley India Pvt. Ltd. 10/13

 Machine Learning using Python
test_ds_predicted = nb_clf.predict( test_X.toarray() )

10.3.4 Print clasification report

from sklearn import metrics

print( metrics.classification_report( test_y, test_ds_predicted ) )

precision recall f1-score support

0 0.98 0.97 0.98 873

1 0.98 0.99 0.98 1203

micro avg 0.98 0.98 0.98 2076

macro avg 0.98 0.98 0.98 2076
weighted avg 0.98 0.98 0.98 2076

from sklearn import metrics

cm = metrics.confusion_matrix( test_y, test_ds_predicted )

[Link](cm, annot=True, fmt='.2f' );

10.4 Using TF-IDF Vectorizer

from sklearn.feature_extraction.text import TfidfVectorizer

tfidf_vectorizer = TfidfVectorizer( analyzer=stemmed_words,

max_features = 1000)
feature_vector = tfidf_vectorizer.fit( train_ds.text )
train_ds_features = tfidf_vectorizer.transform( train_ds.text )
features = feature_vector.get_feature_names()

Copyright © 2019 by Wiley India Pvt. Ltd. 11/13

 Machine Learning using Python
from sklearn.naive_bayes import GaussianNB

train_X, test_X, train_y, test_y = train_test_split( train_ds_features,

train_ds.sentiment,
test_size = 0.3,
random_state = 42 )

nb_clf = GaussianNB()
nb_clf.fit( train_X.toarray(), train_y )

GaussianNB(priors=None, var_smoothing=1e-09)

test_ds_predicted = nb_clf.predict( test_X.toarray() )

print( metrics.classification_report( test_y, test_ds_predicted ) )

precision recall f1-score support

0 0.96 0.96 0.96 873

1 0.97 0.97 0.97 1203

micro avg 0.97 0.97 0.97 2076

macro avg 0.97 0.97 0.97 2076
weighted avg 0.97 0.97 0.97 2076

10.5.1 Using N-grams

import nltk
from [Link] import PorterStemmer
# library for regular expressions
import re

stemmer = PorterStemmer()

def get_stemmed_tokens( doc ):

# Tokenize the documents to words
all_tokens = [word for word in nltk.word_tokenize(doc)]
clean_tokens = []
# remove the all characters other than alphabets. It takes a regex for match
ing.
for each_token in all_tokens:
if [Link]('[a-zA-Z]', each_token):
clean_tokens.append(each_token)
# Stem the words
stemmed_tokens = [[Link](t) for t in clean_tokens]
return stemmed_tokens

Copyright © 2019 by Wiley India Pvt. Ltd. 12/13

 Machine Learning using Python
tfidf_vectorizer = TfidfVectorizer(max_features=500,
stop_words='english',
tokenizer=get_stemmed_tokens,
ngram_range=(1,2))

feature_vector = tfidf_vectorizer.fit( train_ds.text )

train_ds_features = tfidf_vectorizer.transform( train_ds.text )
features = feature_vector.get_feature_names()

/Users/manaranjan/anaconda/lib/python3.5/site-packages/sklearn/featu
re_extraction/[Link]: UserWarning: Your stop_words may be incon
sistent with your preprocessing. Tokenizing the stop words generated
tokens ['abov', 'afterward', 'alon', 'alreadi', 'alway', 'ani', 'ano
th', 'anyon', 'anyth', 'anywher', 'becam', 'becaus', 'becom', 'befo
r', 'besid', 'cri', 'describ', 'dure', 'els', 'elsewher', 'empti',
'everi', 'everyon', 'everyth', 'everywher', 'fifti', 'formerli', 'fo
rti', 'ha', 'henc', 'hereaft', 'herebi', 'hi', 'howev', 'hundr', 'in
de', 'latterli', 'mani', 'meanwhil', 'moreov', 'mostli', 'nobodi',
'noon', 'noth', 'nowher', 'onc', 'onli', 'otherwis', 'ourselv', 'per
hap', 'pleas', 'seriou', 'sever', 'sinc', 'sincer', 'sixti', 'someo
n', 'someth', 'sometim', 'somewher', 'themselv', 'thenc', 'thereaf
t', 'therebi', 'therefor', 'thi', 'thu', 'togeth', 'twelv', 'twent
i', 'veri', 'wa', 'whatev', 'whenc', 'whenev', 'wherea', 'whereaft',
'wherebi', 'wherev', 'whi', 'yourselv'] not in stop_words.
sorted(inconsistent))

10.5.2 Build the model using n-grams

train_X, test_X, train_y, test_y = train_test_split( train_ds_features,

train_ds.sentiment,
test_size = 0.3,
random_state = 42 )

nb_clf = BernoulliNB()
nb_clf.fit( train_X.toarray(), train_y )
test_ds_predicted = nb_clf.predict( test_X.toarray() )
print( metrics.classification_report( test_y, test_ds_predicted ) )

precision recall f1-score support

0 1.00 0.94 0.97 873

1 0.96 1.00 0.98 1203

micro avg 0.97 0.97 0.97 2076

macro avg 0.98 0.97 0.97 2076
weighted avg 0.97 0.97 0.97 2076

Copyright © 2019 by Wiley India Pvt. Ltd. 13/13