Attention-aware with stacked embedding for sentiment analysis of student feedback through deep learning techniques

Area of Science:

• Natural Language Processing (NLP)
• Artificial Intelligence (AI)

Background:

• Sentiment analysis and polarity prediction are critical for understanding user opinions from public data.
• Existing methods face challenges in accurately assessing sentiments in student feedback and social media comments.
• Text embedding and deep learning models offer potential but require refinement for complex NLP tasks.

Purpose of the Study:

• To address the challenges in automatic polarity prediction and sentiment analysis of student feedback.
• To propose and evaluate an innovative hybrid model for enhanced sentiment analysis.
• To compare the proposed model's performance against existing state-of-the-art deep learning techniques.

Main Methods:

• Developed a hybrid model integrating ensemble learning-based text embedding.
• Incorporated a multi-head attention mechanism for improved feature extraction.
• Utilized a combination of deep learning classifiers for sentiment classification.

Main Results:

• The proposed hybrid model achieved 95% accuracy, 97% recall, 95% precision, and a 96% F1-score.
• Demonstrated superior performance compared to existing state-of-the-art deep learning-based sentiment analysis methods.
• Effectively analyzed sentiment in student feedback, showcasing high predictive power.

Conclusions:

• The novel hybrid model significantly advances sentiment analysis capabilities in NLP.
• The model's high performance metrics validate its effectiveness for analyzing student feedback and user sentiments.
• This approach offers a robust solution for applications requiring accurate polarity prediction and sentiment classification.