Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Research Experiment on Multi-Model Comparison for Chinese Text Classification Tasks

JiaCheng Li

TL;DR

This study benchmarks three approaches for Chinese text classification—TextCNN, TextRNN, and FastText—on the THUCNews dataset to guide model choice in real-world settings. It contrasts traditional features, sequence modeling, and fast embedding-based methods, evaluating using Cross-Entropy Loss, Accuracy, Precision, Recall, and F1-Score. FastText consistently achieves the best overall performance (accuracy around $92.02\%$) while TextCNN and TextRNN offer competitive results with domain-specific strengths. The results underscore the practicality of lightweight, n-gram–enhanced bag-of-words models for large-scale Chinese text classification, delivering strong accuracy with high efficiency and interpretability.

Abstract

With the explosive growth of Chinese text data and advancements in natural language processing technologies, Chinese text classification has become one of the key techniques in fields such as information retrieval and sentiment analysis, attracting increasing attention. This paper conducts a comparative study on three deep learning models:TextCNN, TextRNN, and FastText.specifically for Chinese text classification tasks. By conducting experiments on the THUCNews dataset, the performance of these models is evaluated, and their applicability in different scenarios is discussed.

Research Experiment on Multi-Model Comparison for Chinese Text Classification Tasks

TL;DR

This study benchmarks three approaches for Chinese text classification—TextCNN, TextRNN, and FastText—on the THUCNews dataset to guide model choice in real-world settings. It contrasts traditional features, sequence modeling, and fast embedding-based methods, evaluating using Cross-Entropy Loss, Accuracy, Precision, Recall, and F1-Score. FastText consistently achieves the best overall performance (accuracy around ) while TextCNN and TextRNN offer competitive results with domain-specific strengths. The results underscore the practicality of lightweight, n-gram–enhanced bag-of-words models for large-scale Chinese text classification, delivering strong accuracy with high efficiency and interpretability.

Abstract

With the explosive growth of Chinese text data and advancements in natural language processing technologies, Chinese text classification has become one of the key techniques in fields such as information retrieval and sentiment analysis, attracting increasing attention. This paper conducts a comparative study on three deep learning models:TextCNN, TextRNN, and FastText.specifically for Chinese text classification tasks. By conducting experiments on the THUCNews dataset, the performance of these models is evaluated, and their applicability in different scenarios is discussed.

Paper Structure

This paper contains 23 sections, 5 equations, 2 figures, 4 tables.

Table of Contents

  1. INTRODUCTION
  2. Related Technologies
  3. Classical Models' Application and Optimization
  4. Application of Deep Learning Models
  5. Application of Pre-trained Models
  6. Model Design
  7. Text CNN
  8. TextRNN
  9. FastTEXT
  10. The progress of experiment
  11. Experimental Dataset
  12. Dataset Extraction and Preprocessing
  13. Model Training
  14. Experimental Results and Analysis
  15. Cross-Entropy Loss
  16. ...and 8 more sections

Figures (2)

  • Figure 1: The structure of TextCNN
  • Figure 3: The structure of TextRNN