Fuzzy-Biemm: An Emotion Recognition Model Based on Bidirectional Deep Learning and Extended Fuzzy Markov Model

Abstract

Nowadays, a huge number of messages is transmitted by users on social media such as Twitter, Amazon and Facebook. A lot of data and information are exchanged in these media. Considering the need of these social media to detect the negative or positive feelings of users in the text and news, the idea of opinion mining has been proposed. Opinion mining provides the possibility of analyzing users' opinions and discovering knowledge to detect emotions in social media. Some of the most important challenges in social media have been the lack of accuracy, transparency, and accuracy in detecting users' feelings. Various methods have been proposed to detect the sentiments of users based on opinion mining in social media, which, despite their many applications, still face challenges such as lack of accuracy in sentiment analysis. Therefore, in this paper, a sentiment recognition system called Fuzzy-BiEMM based on extended Markov model (EMM), Bi-LSTM deep neural network and fuzzy logic is proposed. Fuzzy logic approach is used to derive effective rules, Bi-LSTM deep neural network is used for sentiment recognition, EMM is used to improve deep neural network. In this paper, the customer datasets of Amazon, Twitter, Facebook, fake news of Covid-19, Amazon and fake news network are used. By simulating the proposed Fuzzy-BiEMM approach, it was observed that the average accuracy of emotion recognition was 96.75%, which is 7.62% better than the proposed Fuzzy-BiEMM method without applying the fuzzy logic approach and 5.02% better than the CSO-LSTMNN approach.