The disappearance of ‘old women’ from Korean headlines

Entry type: Single project

Country/area: South Korea

Publishing organisation: The Kyunghyang Shinmun

Organisation size: Big

Publication date: 2022-07-07

Language: Korean

Authors: Jihye Shin, Soomin Lee, Arum Lee, Hyungkook Jo


Jihye Shin who is a journalist of the digital editing team, responsible for editing and formatting digital content for publication.
Lee soomin who is a journalist of the data journalism team, responsible for using data analysis techniques to produce news articles and stories, and publishing interactive news.
Lee arum who is a journalist of the new contents team, responsible for researching and developing new and innovative content for the media company, and editing content for publication.
Jo hyungkook is a journalist of the data journalism team, responsible for using data analysis techniques to produce news articles and stories.

Project description:

Our analysis of 7.63 million headlines from the top 10 daily newspapers in Korea over the past 10 years, using machine learning, has detected a change in the way the Korean media portrays women. We were able to detect changes in sexism expressions and in terms related to feminism and gender. It puts meaning on positive change, but also contains the opinions of readers in different age groups and the views of educators in the field.

Impact reached:

“A backward step is memorable, but progress is not noticeable. It’s like an article that visualizes it.”
It was an impression that a reader left while posting an article on his SNS. It is memorable because it accurately grasped the message that was intended to be contained in the article. There was a long way to go, but many readers found meaning in tangible changes. Investigation methodology(https://url.kr/mj6efh) on the interactive page also received good reviews.
When I asked readers through the Instagram account of Kyunghyang Shinmun’s women’s narrative archive team ‘flat’, “How do you deal with the question ‘Are you a femi(feminist)?’,” more than 140 answers were returned. The article in the second episode, which pointed out the misogyny in the classroom, also received many responses on SNS.

Techniques/technologies used:

A crawler that we made using python libraries ‘Beautiful Soup’ and ‘Selenium’ browsered and gathered headlines from 10 daily newspapers in Korea over the past 10 years on news anlaysis system named ‘Bigkinds’.
In order to classify headlines about women and headlines about other topics, the study employed the use of the ‘Khan Gender Classifier’ model developed by Junbum Lee who trained Korean comments ELECTRA(KcELECTRA). The model we used is based on the Korean comments ELECTRA (KcELECTRA).
The article utilized Mecab-ko, a morphological analysis tool from the KoNLPy library, to identify the words that were frequently used in headlines about women.
In order to determine whether headlines about women were more emotional than those about other topics, we conducted an emotional analysis. This analysis was performed using the ‘Pororo’ tool from kakaobrain, which is a natural language processing tool that can identify and quantify emotions in text.

Context about the project:

Recent controversies surrounding hate crimes and discrimination against women, including cases such as “Nth Room,” “Female Polices are Useless Theory,” “Seoul Station Assault,” and the “The controversy surrounding the short hair of Ansan who is an Archery athlete” have brought attention. ‘Gender conflicts’ have become a significant source of division in South Korean society. Some political figures have used anti-feminist sentiment to rally supporters, the media expanded the conflicts by mass-producing unconfirmed rumors. The current social atmosphere has made it difficult to address and resolve issues of gender equality and stereotypes. Despite the fact that structural gender discrimination remains a reality in South Korean society, it is difficult to raise awareness about the issue and to initiate a conversation about it.
Despite the backlash, it is important to evaluate efforts to spread gender equality. By confirming the results of efforts for gender equality, we can confirm its direction and try to gauge the direction of gender equality in Korean society.

What can other journalists learn from this project?

The development and advancement of machine learning technology has been widely adopted across various fields, but its application in the media industry in South Korea has been limited. This article shows the potential of using deep learning models and various morpheme analysis libraries for in-depth data analysis in media companies. By creating a separate analysis methodology page, this page would include information on the tools and how they are used and applied to articles. They are expected to be beneficial for journalists who are interested in morpheme analysis.

Project links: