A Classification System Approach in Predicting Chinese Censorship

TL;DR

This study develops and compares classification models, including logistic regression and fine-tuned BERT, to predict Chinese censorship of Weibo posts, demonstrating BERT's superior performance in this task.

Contribution

The paper introduces a new Chinese censorship dataset and evaluates multiple classification approaches, highlighting the effectiveness of fine-tuned BERT for censorship prediction.

Findings

Fine-tuned BERT outperforms other models in accuracy and ROC-AUC.

A new Chinese phrase dataset with censorship labels was created.

Logistic regression models provide a baseline for censorship prediction.

Abstract

This paper is dedicated to using a classifier to predict whether a Weibo post would be censored under the Chinese internet. Through randomized sampling from \citeauthor{Fu2021} and Chinese tokenizing strategies, we constructed a cleaned Chinese phrase dataset with binary censorship markings. Utilizing various probability-based information retrieval methods on the data, we were able to derive 4 logistic regression models for classification. Furthermore, we experimented with pre-trained transformers to perform similar classification tasks. After evaluating both the macro-F1 and ROC-AUC metrics, we concluded that the Fined-Tuned BERT model exceeds other strategies in performance.