Lightweight English Text Classification with Deep Learning Based on Complex System Theory

Area of Science:

• Artificial Intelligence
• Natural Language Processing
• Computational Linguistics

Background:

• Increasing volumes of English text data necessitate efficient classification methods.
• Existing models struggle with generalization in small-sample and cross-domain scenarios.
• Lightweight models are crucial for resource-constrained environments.

Purpose of the Study:

• To develop an efficient English text classification model with improved generalization for few-shot and cross-domain tasks.
• To create a lightweight model that reduces computational cost.
• To leverage complex systems theory and deep learning for text classification.

Main Methods:

• Constructed a graph neural network model incorporating complex systems theory and deep learning.
• Employed a two-level meta-distillation method with meta-learning strategies for knowledge transfer.
• Dynamically adjusted teacher model parameters to optimize the learning process.

Main Results:

• Achieved superior classification performance in few-shot and cross-domain tasks compared to traditional graph neural networks and other models.
• Demonstrated significant reduction in computational time with high performance retention on multi-topic datasets.
• Maintained high efficiency and stable performance in long text processing, outperforming traditional distillation methods.

Conclusions:

• The proposed model offers a feasible and efficient solution for English text classification in resource-constrained settings.
• Effectively enhances classification performance in low-sample and cross-domain scenarios.
• Significantly reduces computational costs, making advanced text classification more accessible.