Sentiment Analysis On Movie Reviews Using Recurrent Neural Network
  • Author(s): Sumesh Kumar Nair ; Ravindra Soni
  • Paper ID: 1700620
  • Page: 242-251
  • Published Date: 29-04-2018
  • Published In: Iconic Research And Engineering Journals
  • Publisher: IRE Journals
  • e-ISSN: 2456-8880
  • Volume/Issue: Volume 1 Issue 10 April-2018
Abstract

In this paper i have done sentiment analysis on IMDB dataset using Recurrent Neural network.Sentiment analysis based on text mining or opinion mining based on different dataset.Sentiment classification is done in three categories- Positive,Neagtive and Neutral.Text classification is done on the dataset and data preprocessing is done to remove hastags, synonms,acronyms etc.LSTM Recurrent Neural Networks to other algorithms for classifying the sentiment of movie reviews.Recurrent neural network provides high accuracy and polarity as compared to different machine learning classifiers.To address this task deep learning has become popular method. LSTM (Long short-term memory) model has been used which is a modified version of RNN (Recurrent Neural Networks). Recurrent Neural Networks has ability handle sequential data very effectively and without performing any feature engineering it can learn directly from low-level features. Instead of exploring LSTMs abilities and capabilities, main focus was to learn how embedding can help us to understand user expectations from text. Proper pre- processing for data has been implemented. Informal language, contextualization, bad grammatical structure, misspellings are additional complicating factors. Reviews are analysed as binary classification task, after processing reviews are classified as either negative or positive. Features for training and testing the deep learning model were retrieved by using new method called ?word-vector?. Moreover, effect of sentence length has also been investigated. Sentiment analysis for short sentences becomes difficult because of lack of contextual information. Multiple hidden layers have been used in the architecture. Dropout, Normalization and Parametric Rectified Linear Unit (PReLU) technology has been used to generalize and improve the accuracy of model. Also, the impact of various hyper-parameter has analysed. Different neural network configurations are evaluated. The performance of model is discussed with respect to the input data and modelconfiguration.

Keywords

Sentiment analysis; PReLU; LSTM; RNN

Citations

IRE Journals:
Sumesh Kumar Nair , Ravindra Soni "Sentiment Analysis On Movie Reviews Using Recurrent Neural Network" Iconic Research And Engineering Journals Volume 1 Issue 10 2018 Page 242-251

IEEE:
Sumesh Kumar Nair , Ravindra Soni "Sentiment Analysis On Movie Reviews Using Recurrent Neural Network" Iconic Research And Engineering Journals, 1(10)