TWEERIFY: A Web-Based Sentiment Analysis System Using Rule and Deep Learning Techniques

Abstract

Social media platforms have ceased to be a platform for interaction and entertainment alone, and they have become a platform where citizens express their opinions about issues that affect them. In recent years, it has become a powerful plat- form where elections are won and lost. Therefore, organizations and governments are increasingly interested in citizen’s views expressed on social media platforms. This research presents a novel approach to carry out aspect-level sentiment anal- ysis of users’ tweets using rule and convolutional neural network (CNN)-based deep learning technique. The rule-based technique was used to detect and extract senti- ments from preprocessed tweets, while the CNN-based deep learning technique was employed for the sentiment polarity classification. A total of 26,378 tweets collected using “security” and “Nigeria” keywords were used to test the proposed model. The proposed model outperformed existing state-of-the-arts GloVe and word2vec models with an accuracy of 82.31%, recall value of 82.21%, precision value of 82.75% and F1 score of 81.89%. The better performance of the proposed techniques could be as a result of the rule-based techniques that was introduced to capture sentiments expressed in slangs or informal languages which GloVe and word2vec have not been designed to capture

Description

Keywords

Citation

Collections