原典翻譯 - AI translation query http://ctext.org/ zh http://ctext.org/discuss.pl?if=gb&message=8454035 Fri, 20 Mar 2026 00:38:54 GMT http://ctext.org/discuss.pl?if=gb&message=8454035 http://ctext.org/discuss.pl?if=gb&message=6311517 Wed, 18 Mar 2026 16:09:48 GMT http://ctext.org/discuss.pl?if=gb&message=6311517 http://ctext.org/discuss.pl?if=gb&message=7466108 ]]> Tue, 03 Mar 2026 20:26:15 GMT http://ctext.org/discuss.pl?if=gb&message=7466108 http://ctext.org/discuss.pl?if=gb&message=6096152 Tue, 03 Mar 2026 18:05:05 GMT http://ctext.org/discuss.pl?if=gb&message=6096152 http://ctext.org/discuss.pl?if=gb&message=8551562
However the overall process isn't quite as simple as putting text into a model with a well-crafted prompt and directly getting the desired result (though of course that was the first thing that was tried; it does work at least to some degree, and it might become a feasible approach in the future with better and more powerful models). The current process involves a combination of prompted LLM generation together with rule-based and heuristic approaches working in tandem, essentially aiming to "push" the LLM to produce the desired type of output, and constraining/correcting it in ways that tend to help achieve this.
]]> Tue, 03 Mar 2026 17:47:51 GMT http://ctext.org/discuss.pl?if=gb&message=8551562 http://ctext.org/discuss.pl?if=gb&message=6119483 Quote: admin @ 2026-03-01 22:02:19

Translations of premodern Chinese texts generated using Artificial Intelligence have been added to ctext.org. This includes all texts in the pre-Qin and Han corpus that previously lacked translations, the 25 dynastic histories, as well as hundreds of other historical, literary, philosophical and poetic works. Translations will be added for other texts on an ongoing basis.

Translations created using AI will inevitably contain mistakes; all translations are aligned with the source text at the sentence level to facilitate identification and correction of errors. Errors can be corrected using the dictionary function, which shows the sentence-by-sentence alignment and provides a direct editing function.

Additionally, editable natural-language summaries of historical entities (people, written works, office titles etc.) in Chinese and English are now supported to supplement existing structured data in the Data Wiki; over 100,000 summaries are currently included, with more to follow.

If you prefer not to have English translations displayed when viewing texts, simply click "None" under "Translation setting" at the top of the page.



Thank you for adding this feature. Is it possible to share which AI model was used to create the new translations, along with a sample of the prompts used?]]> Mon, 02 Mar 2026 02:26:52 GMT http://ctext.org/discuss.pl?if=gb&message=6119483