Multilingual Sentiment Analysis using RNN-LSTM and Neural Machine Translation

Abstract

The growing age of digital presence has made us realize the importance of data and the need to analyze this vast data. This paper addresses the need to classify the vast amount of multilingual text data to perform Sentiment Analysis (SA). Labeled data for multilingual text classification is hard to retrieve, here Neural Machine translation (NMT) is used to propose a labelling system with the help of labelled English dataset which is found in abundance. The proposed model will use Global Vectors (GloVe) as word embeddings which are fed to Recurrent Neural Network-Long short-term memory (RNN-LSTM) to generate context and to classify text in sentiment classes. The major highlight is the labelling system which is used to solve the problem of collecting labelled multilingual text data. The proposed model can also classify sentiment analysis depending on the content of the text, document, or series of text.