Low Fertility Discourse Among South Korean Employees: An Anonymized Dataset of Online Posts and Comments
Description
Posts and comments on low fertility from Blind, an anonymous community of South Korean employees. Data were preprocessed by removing non-standard characters and platform-specific tags. Posts and comments unrelated to childbirth or exhibiting noise were manually filtered, with short comments below the 75th percentile length threshold excluded. After deduplication, the dataset comprises 1,242 posts and 16,968 comments.
Files
Steps to reproduce
We collected data from Blind, a prominent online community for Korean employees that requires users to authenticate their employment via company email. As of February 2023, Blind hosted more than 8 million users, predominantly from Korea's leading conglomerates; about 80% of employees of Samsung, SK, and Hyundai, the three largest Korean conglomerates, and about 90% of employees of the top 1,000 companies by market capitalization use the site. This platform was chosen for its rich, real-time discourse on a variety of professional and personal topics, including childbearing. Using Selenium, a Python-based web scraping tool, we extracted posts and comments that included specific keywords related to childbearing intentions. These keywords were carefully selected to capture discussions directly relevant to fertility, such as “fertility,” "low birthrate," "childbirth reasons," and "family planning." The dataset primarily includes contributions from users aged 20 to 40 years, a demographic critically relevant to studies on fertility intentions. Approximately 74% of the contributors were men, although women typically engaged more deeply with the content, based on time spent on the platform and interaction rates. Initial data preprocessing to standardize the input for analysis involved cleaning the text data from posts and comments to remove emoticons, special characters, and non-standard punctuation. The Blind platform's tagging feature, which allows users to tag specific corporate or thematic boards, was excluded to maintain our focus on the content's narrative quality. Next, by checking the titles, we manually removed 157 posts and 4,773 comments that were unrelated to childbirth or intentionally contentious. To exclude noise comments that were nonsensical or mocking in tone, we sampled 5% of all comments and categorized 360 noise comments. After checking those noise comments, we found that they were generally shorter than meaningful comments, so we removed all 15,893 comments shorter than 32 words, because 32 words was the 75th percentile of noise comment length. Our data preprocessing left us with 1,242 posts and 16,968 comments, excluding duplicates.
Institutions
- Sungkyunkwan University