Enhanced Aquila Optimizer Combined Ensemble Bi-LSTM-GRU With Fuzzy Emotion Extractor for Tweet Sentiment Analysis and Classification

Abstract

Tweet Sentiment Analysis (TSA) is a prominent technique to glean valuable insights from the users of Twitter (currently the X platform) to understand their emotional state of mind. This sentiment and opinion mining from tweets is highly important to formulate a decision-support model in the sector of natural language processing. In this research study, a combined version of the deep learning recurrent model, a fuzzy decision support system, and a variant of the nature-inspired Aquila Optimizer are suggested for sentiment analysis and classification of tweets. Fuzzy Sentiment Emotion Extractor (FEE) is crafted to extract the positive and negative emotions from the tweets using the semantic, positive or negative polarity of the words and their positions. An ensemble Bidirectional Long Short-Term Memory (enBi-LSTM) has been engineered in this work for capturing the order of tweet words, context based semantic detail, and the relation among the words in the tweet sequence. For the proposed ensemble bi-LSTM-GRU models, the weights are assigned to the features with the modeled Enhanced Aquila Optimizer (EAQ), and this intends to attain the prominent features from the tweet word sequence. A combination of the fuzzy emotion extractor and novel ensemble Bi-LSTM-GRU optimized with the enhanced aquila optimization algorithm is evaluated for Twitter (X) datasets. The efficacy of the suggested EAQ-FEE-enBi-LSTM-GRU is proven by its superiority derived from state-of-the-art techniques compared to the same tweet datasets. Evaluated performance metrics for the Twitter datasets claim for the efficacy of the proposed optimized fuzzy emotion extractor-based ensemble Bi-LSTM-GRU model.

Keywords

• Ensemble Bi-LSTM model
• gated recurrent unit
• fuzzy emotion extractor
• fuzzy sentiments
• tweet sentiment analysis
• deep learning