An Intelligent Bi-directional Gate Recurrent Neural Network Based Hybrid Deep Learning Model for Text-based Sentiment Analysis

Sentiment Analysis of social media tweet data involves the interpretation of people’s opinions, emotions, likings, and tendencies expressed in the form of text. Text acts as an essential form of expression written in a language which finds a wide variety of insights for variety of business applications. The major characteristics of texts on social media are randomness, brevity, uncertainty, ambiguity and complexity which pose problems such as feature extraction and polysemy. In order to solve this problem, hybrid deep learning model with fusion attention mechanism for text-based Sentiment analysis has been proposed. The proposed hybrid deep learning model collates Convolutional Neural Networks (CNN) for drawing out the local information and Bi-directional Gated Recurrent Networks (Bi-GRU) to pull the background connection with the text to optimize the emphasis on textual words that have a strong emotional inclination. Social media tweet data from Facebook, Whatsapp, and Twitter are collected and utilized to carry out the experimentation. The research explores semantic sentiment analysis based on the attention mechanism to analyze classification prediction for evaluating both the favorable and detrimental views of the public. Experimental results revealed that the proposed model showed a significant impact in identifying the sentiments of the tweets effective at extracting features from the text and categorizing them than its counterparts with high accuracy.