Text sentiment analysis with CNN & GRU model using GloVe
Abdelhaq Zouzou, Ikram El Azami
Abstract
The sentiment analysis is crucial to understanding people’s position, attitude, and opinion about a given event, which has many applications, such as movie review, advertising, electoral prediction, and evaluation of products. There are several techniques for sentiment analysis, but recently most researchers used word embedding methods in the sentiment classification tasks, word2vec, genism and Global Vector (GloVe) are presently among the best usable and accurate word embedding methods which can transform words on a meaningful vector. In this paper, we propose to use GloVe as a word embedding and introduce a developed classification using convolutional neural network (CNN), Gated Recurrent Unit (GRU), and a hybrid model of GRU and CNN applied on IMDB consist of 50k movie review, then we used Adadelta and Adam optimizer. Experimental results show that the CNN_GRU model with the Adadelta optimizer function achieved good classification results with 86.34% on training accuracy value.