How Chosun Ilbo Built a Scalable English News Translation Pipeline with Solar Pro

Chosun Ilbo is one of South Korea’s largest news organizations, publishing more than a thousand articles every day for domestic and international readers. Expanding its global audience was a strategic priority, but English article production relied heavily on manual translation—resulting in bottlenecks in speed, quality, and consistency.

To overcome these limitations, Chosun Ilbo built an automated translation pipeline powered by Solar Pro, enabling large-scale, newsroom-grade English content production.

‍

1. Bottlenecks in Throughput and Quality Consistency

English articles were primarily translated and reviewed by reporters and editors.
As a result:

• Only a small fraction of Korean articles were available in English.
• Producing timely translations for breaking news or feature stories was difficult.
• Stylistic inconsistency occurred because each translator had different preferences.

2 Literal Translations that Didn’t Fit Newsroom Standards

Conventional machine translation struggled with context and nuance, often producing awkward, overly literal results. Headlines were especially problematic.

Korean original
대통령실 "핵잠수함 큰 줄거리 합의… 앞으로 협의는 많이 해야"

Typical machine translation
Presidential Office Says "Agreed on Big Picture of Nuclear Submarine… Need to Do a Lot of Consultation Going Forward"

The complex Korean structure is carried over verbatim, resulting in an overly long, unnatural English headline that doesn’t follow journalistic style.

3 Difficulty Handling Proper Nouns and Cultural References

Conventional MT also struggled with:

• Film titles requiring creative English equivalents
• Romanization inconsistencies for Korean names (e.g., Sung vs. Seong)
• Applying newsroom-specific style guides or internal glossaries

Because of this, extensive post-editing was required, limiting the ability to scale English news production.

Chosun Ilbo needed a system that could reliably handle context, terminology, and journalistic style—at scale.

‍

Chosun Ilbo built a translation pipeline that decomposes an article into predefined units, applies glossary-based normalization, and generates context-aware translations, including English-style headlines.

1. Article Segmentation

Korean source articles are separated into six structured JSON fields the newsroom already uses:

• Main body
• Title & subtitle
• SEO text
• Captions
• External quotes
• Byline and metadata

2. Glossary Enforcement

The pipeline applies Chosun Ilbo’s official terminology list to ensure:

• Consistent romanization of names
• Unified organization and institution names
• Proper handling of film titles or TV programs that shouldn’t be translated literally

3. Context-Aware Full-Article Translation

Solar Pro performs translation on the entire article at once—six sections in a single pass—to prevent sentence-level context loss and produce natural, newsroom-appropriate English.

4. Natural English Headline Generation

Instead of directly translating Korean headlines, Solar Pro generates a new English headline based on article content and English news style conventions.

Korean title
대통령실 “핵잠수함 큰 줄거리 합의… 앞으로 협의는 많이 해야”

Generated English title
South Korea, U.S. Agree on Nuclear Submarine Framework

5. Structured JSON Output

Translated results are returned in structured form so the newsroom CMS can ingest them immediately.

This end-to-end system eliminates repetitive corrections and enforces terminology and stylistic consistency across all English articles.

‍

The results were immediate and transformative.

Key Metrics

• ~30× increase in translation volume:
  Without hiring additional staff, Chosun Ilbo can now publish English versions of a vastly larger share of its Korean articles.
• 10× growth in English pageviews:
  Higher translation throughput and improved quality boosted global visibility and readership.

The automation has laid the foundation for Chosun Ilbo to operate as a truly global news publisher.

‍

Testimonial

“Our collaboration with Upstage has become a major turning point in Chosun Ilbo’s global news service strategy. Compared to AI models from global big tech companies, Solar clearly demonstrated the strongest capability in understanding Korean. We look forward to continuing the leadership we have built together in translation-AI technology.”
— Gunhyung Park, Head of Content & AI Strategy, Chosun Ilbo

Before Upstage

• Majority of English articles produced manually
• Only a small subset of Korean content available in English
• Frequent context and terminology issues requiring re-editing
• Inconsistent romanization and proper-noun handling
• Literal, unnatural English headlines

After Upstage

• Fully automated workflow: segmentation → glossary → translation → headline
• ~30× more articles translated
• 10× increase in English pageviews
• Standardized terminology and romanization across articles
• Natural, newsroom-grade English headlines generated automatically

‍