a method receives a first translation for an instance of content.
In some embodiments, a method receives a first translation for an instance of content. A first prompt is generated to refine the first translation. The first prompt includes instructions to refine the first translation and a scoring guide. The method inputs the first prompt into a large language model to generate a score for the first translation and suggestions for the first translation. The method evaluates the score to determine whether the first translation meets a threshold. When the score does not meet the threshold: the method generates a second prompt to refine the first translation, inputs the second prompt into the large language model to refine the first translation using one or more of the suggestions to generate a second translation, and provides feedback to the generate another first prompt to refine the second translation. When the score meets the threshold, the method outputs the first translation.
BACKGROUND
A content delivery system may release content in multiple different countries. This may require translation of the content to multiple languages. The translations may use a large amount of resources and require a large amount of time. A machine translation may be used to reduce the translation time. However, the translation of media content may require integrating localized linguistic styles, which may not be captured in the machine translations. Also, the machine translations may lack human naturalness, such as the translation may be stiff, robotic, overly literal, and may not be appropriate for translations of media content.
1. A method comprising: receiving a first translation for an instance of content; generating a first prompt to refine the first translation, wherein the first prompt includes instructions to refine the first translation and a scoring guide; inputting the first prompt into a large language model to generate a score for the first translation and suggestions for the first translation; evaluating the score to determine whether the first translation meets a threshold; when the score does not meet the threshold: generating a second prompt to refine the first translation using the suggestions; inputting the second prompt into the large language model to refine the first translation using one or more of the suggestions to generate a second translation; providing feedback to generate another first prompt to refine the second translation using the large language model; and when the score meets the threshold, outputting the first translation. ||
17. A non-transitory computer-readable storage medium having stored thereon computer executable instructions, which when executed by a computing device, cause the computing device to be operable for: receiving a first translation for an instance of content; generating a first prompt to refine the first translation, wherein the first prompt includes instructions to refine the first translation and a scoring guide; inputting the first prompt into a large language model to generate a score for the first translation and suggestions for the first translation; evaluating the score to determine whether the first translation meets a threshold; when the score does not meet the threshold: generating a second prompt to refine the first translation using the suggestions; inputting the second prompt into the large language model to refine the first translation using one or more of the suggestions to generate a second translation; providing feedback to generate another first prompt to refine the second translation using the large language model; and when the score meets the threshold, outputting the first translation. ||
20. An apparatus comprising: one or more computer processors; and a computer-readable storage medium comprising instructions for controlling the one or more computer processors to be operable for: receiving a first translation for an instance of content; generating a first prompt to refine the first translation, wherein the first prompt includes instructions to refine the first translation and a scoring guide; inputting the first prompt into a large language model to generate a score for the first translation and suggestions for the first translation; evaluating the score to determine whether the first translation meets a threshold; when the score does not meet the threshold: generating a second prompt to refine the first translation using the suggestions; inputting the second prompt into the large language model to refine the first translation using one or more of the suggestions to generate a second translation; providing feedback to generate another first prompt to refine the second translation using the large language model; and when the score meets the threshold, outputting the first translation.