Computer Organization: National Institute of Technology Hamirpur

National Institute of Technology
Hamirpur
CSD-221
COMPUTER ORGANIZATION
Text Sentiment Predictor
Submitted By :
185010 Akshay Kumar akshaychoudharyac01@gmail.com
185019 Prateek Bharat Sharma prateek21112@gmail.com
185036 Aman Garg amangarg3april@gmail.com
{ B. Tech CSE 4yr }
Submitted To :
Dr. Jatoth Chandrashekhar
Assistant Professor
Computer Science & Engineering Dept.
NIT Hamirpur
Text Sentiment Predictor 185010,185019,185036
Contents
1 Introduction : 2
1.1 LSTM (Long-Short-Term-Memory) . . . . . . . . . . . . . . . . . . . 2
2 Text Sentiment Predictor 3
3 Code : 4
3.1 Train . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
3.2 Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1
1 Introduction :
A recurrent neural network (RNN) is a class of artificial neural networks
where connections between nodes form a directed graph along a temporal sequence.
This allows it to exhibit temporal dynamic behavior. Derived from feedforward neu-
ral networks, RNNs can use their internal state (memory) to process variable length
sequences of inputs. This makes them applicable to tasks such as unsegmented,
connected handwriting recognition[2] or speech recognition.
The term “recurrent neural network” is used indiscriminately to refer to two
broad classes of networks with a similar general structure, where one is finite impulse
and the other is infinite impulse. Both classes of networks exhibit temporal dynamic
behavior.A finite impulse recurrent network is a directed acyclic graph that can be
unrolled and replaced with a strictly feedforward neural network, while an infinite
impulse recurrent network is a directed cyclic graph that can not be unrolled.
1.1 LSTM (Long-Short-Term-Memory)

Long short-term memory (LSTM) is an artificial recurrent neural network
(RNN) architecture.used in the field of deep learning. Unlike standard feedforward
neural networks, LSTM has feedback connections. It can not only process single
data points (such as images), but also entire sequences of data (such as speech or
video). For example, LSTM is applicable to tasks such as unsegmented, connected
handwriting recognition speech recognition and anomaly detection in network traffic
or IDSs (intrusion detection systems).
A common LSTM unit is composed of a cell, an input gate, an output gate
and a forget gate. The cell remembers values over arbitrary time intervals and the
three gates regulate the flow of information into and out of the cell.
LSTM networks are well-suited to classifying, processing and making predic-
tions based on time series data, since there can be lags of unknown duration between
important events in a time series. LSTMs were developed to deal with the vanishing
gradient problem that can be encountered when training traditional RNNs. Rela-
tive insensitivity to gap length is an advantage of LSTM over RNNs, hidden Markov
models and other sequence learning methods in numerous applications.
Figure 1: Standard RNN and Unfolded RNN
2
2 Text Sentiment Predictor

Text Sentiment Predictor is an application that takes input a text message
and outputs an emoji and sentiment for that message. For Example if the input
text message is ”i am not good”, then the application will output ”Guilt” Sentiment
and a Corresponding Emoji for it. This Text sentiment predictor is made using
Long short-term memory (LSTM) architecture of Recurrent Neural Network(RNN).
The model is trained using Adam Optimizer with batch size of 32. The accuracy
achieved by our model is 87%.
The Dependencies of the model are:
• numpy
• pandas
• keras
• tensorflow
• emoji
• nltk
The summary of various layers of trained model is as follows:
Layer (Type) Output Dimension Trainable Parameters
input1 (InputLayer) (None, 40) 0
embedding1 (Embedding) (None, 40, 300) 120000300
bidirectional1 (Bidirectional) (None, 40, 256) 439296
dropout1 (Dropout) (None, 40, 256) 0
lstm2 (LSTM) (None, 128) 197120
dropout2 (Dropout) (None, 128) 0
dense1 (Dense) (None, 7) 903
activation1 (Activation) (None, 7) 0
Total Parameters : 120,637,619
Trainable Parameters : 637,319
Non-trainable Parameters : 120,000,300
The fully Trained Model can be found on the following link:
Text Sentiment Predictor
https://github.com/aman-garg0001/Text-Sentiment-Predictor
To run the Application first download all the dependencies. Then write input
string in the input.txt and then run the test.py file and the output will be in the
output.txt file.
3
3 Code :
The Code for our train and test file are:
3.1 Train
Listing 1: train.py
1 import pandas as pd
2 import numpy as np
3 import emoji
4
5 data_set = pd . read_csv ( ’ emoji_data . csv ’ , engine = ’ python ’)
6
7 text = data_set [ ’ Text ’]
8 label = data_set [ ’ Label ’]
9
10 emoji_dict = { " joy " : " " , " fear " : " " , " anger " : " " , " sadness " : " " ,←-
" disgust " : " " , " shame " : " " , " guilt " : " " }
11 emoji = { " joy " :0 , " fear " :1 , " anger " :2 , " sadness " :3 , " disgust←-
" :4 , " shame " :5 , " guilt " :6}
12
13 for i in range ( len ( label ) ) :
14 label [ i ] = emoji [ label [ i ]]
15
16 Y = np . zeros (( len ( label ) , 7) )
17 for i in range ( len ( label ) ) :
18 Y [i , label [ i ]] = 1
19
20 def read_glove_vecs ( glove_file ) :
21 with open ( glove_file , ’r ’) as f :
22 words = set ()
23 word_to_vec_map = {}
24 for line in f :
25 line = line . strip () . split ()
26 curr_word = line [0]
27 words . add ( curr_word )
28 word_to_vec_map [ curr_word ] = np . array ( line [1:] , ←-
dtype = np . float64 )
29
30 i = 1
31 words_to_index = {}
32 index_to_words = {}
33 for w in sorted ( words ) :
34 words_to_index [ w ] = i
35 index_to_words [ i ] = w
36 i = i + 1
37 return words_to_index , index_to_words , word_to_vec_map
38
4
39 ! wget http :// nlp . stanford . edu / data / glove .6 B . zip

40 ! unzip glove .6 B . zip
41
42 words_to_index , index_to_words , word_to_vec_map = ←-
read_glove_vecs ( ’ glove .6 B .300 d . txt ’)
43 from keras . models import Model
44 from keras . layers import Dense , Input , Dropout , LSTM , ←-
Activation , Bidirectional
45 from keras . layers . embeddings import Embedding
46 from keras . preprocessing import sequence
47 from keras . initializers import glorot_uniform
48
49 from nltk . tokenize import TweetTokenizer
50
51 def s e n t e n c e s _ t o _ i n d i c e s (X , words_to_index , max_len ) :
52 m = X . shape [0]
53 X_indices = np . zeros (( m , max_len ) )
54 tknz = TweetTokenizer ()
55 for i in range ( m ) :
56 sentence_words = [ word . lower () for word in tknz .←-
tokenize ( X [ i ]) ]
57 j = 0
58 for w in sentence_words :
59 if j > 39:
60 break
61 try :
62 X_indices [i , j ] = words_to_index [ w ]
63 j = j + 1
64 except KeyError :
65 continue
66 return X_indices
67
68 X_set = np . array ( text )
69 max_len = 40
70 X_set = s e n t e n c e s _ t o _ i n d i c e s ( X_set , words_to_index , max_len )
71
72 vocab_len = len ( words_to_index ) + 1
73 emb_dim = word_to_vec_map [ " cucumber " ]. shape [0]
74 emb_matrix = np . zeros (( vocab_len , emb_dim ) )
75 for word , idx in words_to_index . items () :
76 emb_matrix [ idx , :] = word_to_vec_map [ word ]
77 embedding_layer = Embedding ( vocab_len , emb_dim , trainable = ←-
False )
78 embedding_layer . build (( None ,) )
79 embedding_layer . set_weights ([ emb_matrix ])
80
81 sentence_indices = Input (( max_len , ) , dtype = ’ int32 ’)
82
83 embeddings = embedding_layer ( sentence_indices )
5
84
85 X = Bidirectional ( LSTM ( units = 128 , return_sequences = True )←-
) ( embeddings )
86
87 X = Dropout ( rate = 0.4) ( X )
88
89 X = LSTM ( units = 128 , return_sequences = False ) ( X )
90
91 X = Dropout ( rate = .4) ( X )
92
93 X = Dense ( units = 7) ( X )
94
95 X = Activation ( ’ softmax ’) ( X )
96
97 model = Model ( inputs = sentence_indices , outputs = X )
98
99 model . compile ( loss = ’ c a t e g o r i c a l _ c r o s s e n t r o p y ’ , optimizer = ’←-
adam ’ , metrics =[ ’ accuracy ’ ])
100 from sklearn . model_selection import train_test_split
101 X_train , X_test , y_train , y_test = train_test_split ( X_set , Y←-
, test_size =0.10)
102
103 model . fit ( X_train , y_train , epochs = 20 , batch_size = 32 , ←-
shuffle = True , validation_data =( X_test , y_test ) )
104
105 model . save ( ’ Emoji - Predictor . h5 ’)
3.2 Test
Listing 2: test.py
1 import pandas as pd
2 import numpy as np
3 import emoji
4 from nltk . tokenize import TweetTokenizer
5 import keras
6 import tensorflow
7
8 emoji_dict = { " joy " : " " , " fear " : " " , " anger " : " " , " sadness " : " " ,←-
" disgust " : " " , " shame " : " " , " guilt " : " " }
9 emoji = { " joy " :0 , " fear " :1 , " anger " :2 , " sadness " :3 , " disgust←-
" :4 , " shame " :5 , " guilt " :6}
10
11 def preprocess (X , words_to_index , max_len ) :
12 X_indices = np . zeros ((1 , max_len ) )
13 tknz = TweetTokenizer ()
14 sentence_words = [ word . lower () for word in tknz . tokenize←-
(X)]
6
15 j = 0
16 for w in sentence_words :
17 if j > 39:
18 break
19 try :
20 X_indices [0 , j ] = words_to_index [ w ]
21 j = j + 1
22 except KeyError :
23 continue
24 return X_indices
25
26 max_len = 40
27 words_to_index , index_to_words , word_to_vec_map = ←-
read_glove_vecs ( ’ glove .6 B .300 d . txt ’)
28
29 file = open ( ’ input . txt ’ , ’r ’)
30 inp = file . read ()
31 file . close ()
32 x = preprocess ( inp , words_to_index , max_len )
33 model = keras . models . load_model ( ’ Emoji - Predictor . h5 ’)
34
35 y = model . predict ( x )
36 y = np . argmax (y , axis = -1)
37 arr = list ( emoji_dict . keys () )
38
39 ans = " Predicted Sentiment : " + arr [ y [0]]
40 ans += " \ nPredicted Emoji : " + emoji_dict [ arr [ y [0]]]
41
42 file = open ( ’ output . txt ’ , ’w ’)
43 file . write ( ans )
44 file . close ()

Computer Organization: National Institute of Technology Hamirpur

Uploaded by

Copyright:

Available Formats

Computer Organization: National Institute of Technology Hamirpur

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Computer Organization: National Institute of Technology Hamirpur

Uploaded by

Copyright:

Available Formats

National Institute of Technology

Text Sentiment Predictor

2 Text Sentiment Predictor 3

1.1 LSTM (Long-Short-Term-Memory)

Figure 1: Standard RNN and Unfolded RNN

2 Text Sentiment Predictor

39 ! wget http :// nlp . stanford . edu / data / glove .6 B . zip

You might also like