Optimized hierarchical CLSTM model for sentiment classification of tweets using boosted killer whale predation...

Area of Science:

• Natural Language Processing
• Machine Learning
• Computational Linguistics

Background:

• Social media generates vast user-generated content, increasing the complexity of opinion mining.
• Existing sentiment analysis (SA) systems struggle with prediction accuracy due to data limitations and complex model configurations, hindering deep learning (DL) applications.
• Accurate SA is crucial for understanding public opinion on products, advancements, and societal issues.

Purpose of the Study:

• To develop an improved sentiment analysis model for classifying emotions in tweets.
• To address the limitations of current SA systems, particularly low accuracy and prediction rates in deep learning models.
• To enhance the hierarchical extraction of local emotions from textual data.

Main Methods:

• Proposed an Optimal Tiered Convolutional Neural Long Short-Term Memory (OTCNLSTM) model with classification learning for emotion recognition.
• Utilized four training blocks within the TCNLSTM model for hierarchical local feature extraction.
• Implemented the Boosted Killer Whale Predation (BKWOP) strategy for hyperparameter optimization and stable neural network model construction.
• Conducted comparative experiments using the Kaggle Twitter dataset to evaluate model performance.

Main Results:

• The OTCNLSTM model demonstrated superior performance in classifying tweet emotions compared to other sentiment analysis models.
• The proposed model effectively extracts local emotions hierarchically.
• The BKWOP strategy successfully identified optimal hyperparameters for building a stable neural network.

Conclusions:

• The OTCNLSTM model offers a significant advancement in sentiment analysis accuracy for social media data.
• The hierarchical feature extraction approach enhances the model's ability to recognize nuanced emotions in text.
• This research provides a more robust and accurate solution for real-time commercial sentiment analysis applications.