Summarizer fails on Korean article

leftium

I tried to summarize this Korean article with the Universal Summarizer: https://m.news.nate.com/view/20250704n05464

https://kagi.com/summarizer?target_language=DOC&summary=summary&url=https%3A%2F%2Fm.news.nate.com%2Fview%2F20250704n05464

It resulted in this gibberish:

Title: 6 27 å ? Ʈ, ŷ ҡ '¡ ' : Ʈ

The document presents a numerical and analytical overview related to certain data, possibly in a financial or statistical context. It mentions a percentage (85%) related to a specific metric and discusses various numerical values associated with different categories or items, such as 30 and 28. It also notes the presence of 177 and other figures indicating a structured analysis of data.

The text includes references to various identifiers and possibly classifications, such as α, ǡ, and ϱ, suggesting a categorization of entities or data points. It appears to detail changes or trends over time, with specific focus on percentages and counts that contribute to a broader analysis.

Additionally, the document contains links to related articles, indicating it may be part of a larger news or analytical platform. The overall tone suggests a formal presentation of data, possibly aimed at a professional or academic audience interested in detailed statistical information.

I expected an English summary of this article about the Korean real estate market.

Here is English discussion related to this article:

The "Time saved reading: 1m" seemed a little short (1367 tokens)

Additionally, when I selected output language "Document language" the output was not in Korean as expected, but some language like Portuguese.

Kagi translate eventually worked (there was a DNS error during the first attempt.) It would be nice to be able to then summarize that translated content. (This would be a work-around for the Univeral Summarizer not working.)

Bodine

It seems to also cause issues when translating different articles from the same site. I wonder if the site itself has some anti-web scraping feature to prevent people from training models on its data.

Hanbyeol

You can read the same article on the official website of the media or on Naver, our country's largest "web portal".

I have tested both URLs, and the Summarizer reads them well.

And here's the official press release from our government.
(Sharing this in case OP is curious about the policy.)
https://www.korea.kr/news/policyNewsView.do?newsId=148945077

Hanbyeol

On a side note, it's somewhat surprising to see a random Redditor linking to Nate.com.
In Korea, I don't think anyone really goes there just to read the news. 🤷
It’s kind of like linking to AOL.com or Yahoo.com in the US in 2025, if that makes sense.

The main reason Koreans might visit Nate is to install the messenger app called "NateOn":
https://nateonweb.nate.com/
(NateOn displays a news headline ticker, and "news.nate.com" provides such news articles.)

Some workplaces still use it out of habit from before the days of Teams or Slack.
On mobile, almost everyone now uses another app called "KakaoTalk," which is from a different company.

According to this news article (March 2025), Nate's market share has already dropped to 0%:
https://n.news.naver.com/article/011/0004456731

Direct quote from the article (in Korean): "네이트 0%대 추락"
(Naver holds about 66.61%, and Google about 25.93%.)

So, I don't think Nate is a particularly critical website, and the Kagi team probably doesn't need to worry about it too much. As I wrote just above, if something is published on Nate, it will also be available on Naver.