You are on page 1of 4

Sentiment Analysis

Project
Vasanth Pandiarajan
Parinitha Ganapathisubramaniam Kannan
Pre-processing Steps and Choosing Features
For the Algorithm

Stop Word Removal


Punctuations symbol removal
Conversion of text to lowercase
Word Tokenization
Choosing context window of words as features
Encoding words using pre-trained Word2Vec model
Algorithm Used

 The encoded feature vectors were given as input to a Gated


Recurrent Neural Network
 Each word vector is an input at each time step
 Fully connected layer and softmax applied to output of final
time step
 Class having highest probability was chosen as the label for
the input sentence
 Softmax cross entropy loss and Adam optimizer used
Results Obtained

You might also like