Emotion Analysis Model of Microblog Comment Text Based on CNN-BiLSTM

Area of Science:

• Natural Language Processing
• Artificial Intelligence
• Computational Linguistics

Background:

• Traditional emotion analysis methods rely heavily on manual labor and exhibit limited generalization.
• Existing dictionary-based and machine learning approaches face challenges in accurately capturing nuanced emotions in text.

Purpose of the Study:

• To propose a novel deep learning model for microblog comment text emotion analysis.
• To address the limitations of traditional methods by leveraging advanced neural network architectures.
• To improve the accuracy and generalization capabilities of emotion detection in social media text.

Main Methods:

• Data acquisition via microblog crawler, followed by preprocessing (cleaning, word segmentation, stop word removal).
• Word vector training using the Skip-gram model on a large-scale unlabeled corpus.
• Implementation of a hybrid Convolutional Neural Network (CNN) and Bidirectional Long-Short Term Memory (BiLSTM) model for text input, optimized with Adamax.

Main Results:

• The proposed CNN-BiLSTM model achieved a high accuracy of 0.94 for text emotion analysis.
• Demonstrated an improvement of 8.51% in accuracy compared to a single CNN model.
• Effectively combined the semantic learning of BiLSTM and feature extraction of CNN for enhanced performance.

Conclusions:

• The deep learning-based emotion analysis model significantly outperforms traditional methods.
• The integration of CNN and BiLSTM architectures provides a robust approach for analyzing microblog comment sentiment.
• The model offers a scalable and accurate solution for automated emotion detection in large volumes of social media data.