Natural Language Processing for the Analysis Sentiment using a LSTM Model

Abstract

Over the past decade, social networks have revolutionised the communication between organisations and their customers, and the data provided by customers on social network platforms is having an increasingly important impact on how organisations collect and analyse this data to make better decisions. We have prepared a new dataset that will allow the scientific community to estimate and evaluate new models using nearly the same conditions. Moreover, this dataset represents a recent and interesting sample for the proposed machine learning models to correctly identify the topics or points on which the company should focus to improve customer satisfaction and better meet their needs. Therefore, we have proposed a recurrent neural network (RNN) with Long short-term memory (LSTM) that we will run in the cloud to predict sentiment analysis. The objective is also to define systems capable of extracting subjective information from natural language texts, such as feelings and opinions, with the aim of creating structured knowledge that can be used by a decision support system or a decision maker for better customer management. The proposed neural network has been trained on the proposed dataset which contains 50 000 customer observations. The performance of the proposed architecture is very important as the success rate is 96%.