Development and Evaluation of a KoBERT-Based Topic Classification Model for Political News Articles
Keywords: Political News, KoBERT, Topic Classification, Deep Learning, NLP, Media Frame, Automatic Classification
Submission Type: Abstract
Status: In Review | Submitted at: 2025-07-01 22:27:17
Abstract
This study aims to develop and evaluate a topic classification model for political news articles using KoBERT, a Korean language variant of BERT. We constructed a dataset of news articles mentioning major political figures from the past three years and fine-tuned a KoBERT model to classify the articles into four categories: policy, controversy, election, and image-building. The model achieved an accuracy of 82% and an F1-score of 0.81, outperforming baseline models such as LSTM and Naive Bayes. The results demonstrate the effectiveness of KoBERT for nuanced political text classification and suggest its applicability for further studies on political discourse analysis, media bias detection, and public opinion monitoring.
Authors
- Rackjune Baek (First Author), CKU Professor – rj100@hanmail.net