Wikipedia Translation Guidelines for LLM and Machine Translation
Translating content for the English Wikipedia requires a balance between efficiency and accuracy. While modern tools like Machine Translation (MT)—automated software that translates text from one language to another—can accelerate the process, they cannot replace human oversight. To maintain the encyclopedia's standards, contributors must follow specific protocols when using tools like DeepL or Google Translate.
The Role of Machine Translation and LLMs
Machine translation and Large Language Models (LLMs) serve as useful starting points for creating new articles. However, contributors are strictly prohibited from simply copy-pasting machine-translated text into the English Wikipedia. Instead, editors must adhere to the LLM translation guideline, revising errors and confirming that the final output is accurate and natural.
[ไม่มีภาพประกอบ]
Quality Control and Verification
Not all source material is suitable for translation. Editors should exercise caution and avoid translating text that appears unreliable or is of low quality. To ensure the integrity of the information, it is recommended to verify the claims in the foreign-language article using the provided references before proceeding with the translation.
Categorization and Organization
Proper categorization helps users find content more easily. With 1,457 articles already existing in the main category, contributors are encouraged to specify a topic using the | topic = parameter in the template to aid in precise categorization.
Copyright and Attribution Requirements
Because translations are derivative works, providing proper copyright attribution is mandatory. This must be done in the edit summary accompanying the translation by including an interlanguage link to the original source.
For example, a model attribution summary would read: "Content in this edit is translated from the existing Japanese Wikipedia article at [[:ja:日本経済新聞]]; see its history for attribution."
Additionally, editors may add the {{Translated|ja|日本経済新聞}} template to the article's talk page to further document the source.
Key Facts
- No Direct Copy-Pasting: Machine-translated text must be revised and verified for accuracy.
- Verification: Unreliable or low-quality text should not be translated; use references to verify facts.
- Attribution: An interlanguage link to the source must be included in the edit summary.
- Categorization: Use the
| topic =field to help organize content within the main category of 1,457 articles. - Documentation: The
{{Translated}}template can be used on the talk page.
| Requirement | Action Needed | Purpose |
|---|---|---|
| Tool Usage | Revise LLM/MT output | Ensure accuracy and readability |
| Source Quality | Verify with references | Prevent low-quality content |
| Attribution | Interlanguage link in edit summary | Copyright compliance |
| Categorization | Specify | topic = |
Improve article discoverability |
Frequently Asked Questions
Can I use Google Translate or DeepL for Wikipedia articles?
Yes, but only as a starting point. You must follow the LLM translation guidelines, revise any errors, and confirm the accuracy of the text rather than copy-pasting it directly.
What should I do if the source text seems unreliable?
You should not translate text that appears to be low-quality or unreliable. If possible, verify the information using the references provided in the original foreign-language article.
How do I properly attribute a translation?
You must provide a copyright attribution in your edit summary by including an interlanguage link to the source article and directing users to its history for full attribution.
Is the {{Translated}} template mandatory?
While the edit summary attribution is required, you may also optionally add the {{Translated}} template to the article's talk page for additional clarity.
Why is specifying a topic important for categorization?
Because the main category already contains 1,457 articles, specifying a topic via the | topic = parameter helps organize the content more effectively for readers.