Microblog User Emotion Analysis Method Based on Improved Hierarchical Attention Mechanism and BiLSTM

Area of Science:

• Natural Language Processing
• Computational Linguistics
• Artificial Intelligence

Background:

• Chinese fine-grained emotion analysis aims to identify target words and their emotional polarity within sentences.
• Current Sina Microblog emotion analysis methods suffer from low accuracy, hindering effective prediction and management.

Purpose of the Study:

• To propose an improved emotion analysis method for Sina Microblog users.
• To enhance the accuracy and effectiveness of fine-grained emotion prediction and management.

Main Methods:

• A novel method integrating Bidirectional Long Short-Term Memory (BiLSTM) and an improved hierarchical attention mechanism.
• Utilizing weighted word vectors of TF-IDF to address the dimensionality curse of one-hot representation.
• Employing text-level, subjective, and objective analysis for comprehensive emotion modeling.

Main Results:

• The proposed method achieved superior performance compared to two other methods across 7 emotion classifications.
• Achieved highest precision (95.8%), recall (95.9%), and F1-score (96.1%).
• Demonstrated significantly better algorithm performance and excellent overall model effectiveness.

Conclusions:

• The proposed BiLSTM with an improved hierarchical attention mechanism offers a robust solution for Chinese fine-grained emotion analysis.
• The method effectively enhances context acquisition and focuses on critical emotional elements for improved accuracy.
• This approach provides a promising direction for more accurate Sina Microblog user emotion analysis and management.