Sentiment Analysis using Deep Learning on GPU

Abstract

Sentiment analysis is performed using deep learning approach on airline tweet dataset on GPU and CPU. The tweet airline dataset is downloaded from internet. In word embedding technique, we represent the text with word vector by mapping tweet's token to already pre-trained word vector like APNews corpus. We split the word vector airline dataset into training and testing dataset and build the proposed model. We feed such word vectors to deep learning model like convolutional neural network and analyze the given tweet as either positive or negative opinion. The proposed model is trained using training dataset and trained model is used to validate the testing dataset on GPU and CPU. The experiment on GPU is helpful for parallel and speedup computing. We got the training accuracy 98% and testing accuracy 90% of airline tweet dataset on GPU.