Скачать или смотреть Misinformation in Crime News: Countering Xenophobic Stereotypes [APA TOPSS 2025 Contest Lee.]

I did two analyses: quantitative (statistical) approach and qualitative (Natural Language Processing, NLP) approach.

For the first approach, the data was collected using the Korean News Archive "Big KINDS". I analyzed social crime articles from seven major media outlets, focusing on five major crimes: murder, robbery, arson, sexual violence, and assault. Of the five major crime articles, those involving non-Korean perpetrators were categorized as foreign crimes. Articles explicitly mentioning “foreigner," “migrant,” or a specific nationality were included.

For the second approach, I collected articles through web scraping and then preprocessed them (tokenization, stopword removal, stemming, etc.). Subsequently, I classified the articles’ composition of sentiments into positive, negative, and neutral sentiments, using NLP techniques with three NLP models (kc-BERT, kcELECTRA, Ko-SRoBERTa) to classify article sentiment (positive, negative, neutral).