April 17, 2021ACM Computing Surveys
Deep Learning--based Text Classification
Authors
Discussion
Loading...
Member takes
Implication
Systematic review demonstrates superior performance of deep learning architectures across benchmark natural language tasks, highlighting core strengths and emerging research frontiers.
Key Points
- To review recent advances in deep learning for text classification, analyze the technical characteristics of leading architectures, and benchmark their empirical performance.
- Reviewed more than 150 deep learning-based text classification models to evaluate architectural designs, technical contributions, and similarities.
- Summarized and cataloged over 40 popular text classification benchmark datasets across domains such as sentiment analysis, news categorization, and natural language inference.
- Conducted quantitative performance evaluations across standardized text classification benchmarks.
- Deep learning models consistently outperform classical machine learning algorithms across diverse natural language processing benchmarks.
- Identified critical architectural strengths and common design patterns across 150+ neural network models, establishing a standardized framework for future text classification research.
Cite This Study
Minaee et al. (2021) studied this question.