26,897 news articles from seven major South Korean outlets, collected between January 2013 and April 2021. The corpus focuses on coverage of particulate matter (PM), known as 'mise meonji' in Korean, following the WHO's 2013 classification of PM as a carcinogen. The dataset was created by Youk, Sungbin and is hosted on Harvard Dataverse.
Use Cases
- Analyze media framing and sentiment trends related to air pollution based on the news article corpus.
- Study the evolution of public discourse on environmental health risks based on the nine-year time span.
- Train or fine-tune NLP models for topic modeling or entity recognition on Korean news text based on the collected articles.
- Compare coverage differences across media outlets with varying political or editorial stances based on the seven selected sources.
Strengths
- Substantial corpus of 26,897 articles for longitudinal analysis.
- Clear temporal coverage from January 2013 to April 2021, spanning over nine years.
- Articles sourced from seven leading South Korean news outlets, selected based on circulation and popularity metrics.
- Data collection methodology is described, including keyword search and filtering for relevance.
Limitations
- Column-level documentation is absent; field semantics must be inferred after download.
- Row count is unknown, which may limit suitability assessment for specific modeling tasks.
- Data may reflect source bias inherent to the selected news outlets and the specific search keyword.
Provenance
- Source
- Harvard Dataverse
- Collection Method
- Articles were retrieved via keyword search ('미세먼지') from official news websites, followed by filtering of advertisements and irrelevant content.
- Time Range
- January 2013 to April 2021
- Freshness
- Last updated 2026-07-10 14:16:04; freshness should be verified.
- Geography
- South Korea