Sign in Subscribe
Machine learning

Patent Application for Translation Quality Assurance Based on Large Language Models Revealed

Strategic Insights

— 2 min read

A patent application for Translation Quality Assurance Based on Large Language Models, filed by inventors JIN, YUAN, LIU, XI XI, Matthiesen, Brian Robert, Sun, Wei, and Wang, Li ping, has been made available online. The application, filed on December 5, 2023, and posted on June 5, 2025, proposes a method, system, and computer program product for assuring the quality of machine translations based on large language models.

According to the inventors, machine translation uses either rule-based or probabilistic machine learning approaches to translate text or speech from one language to another, including the contextual, idiomatic, and pragmatic nuances of both languages. The proposed invention aims to improve the accuracy and reliability of machine translation by using large language models to revise translations that are below a certain confidence threshold.

The patent application describes three aspects of the proposed invention: a method for assuring quality of machine translations based on large language models, a computer system for assuring quality of machine translations based on large language models, and a computer program product for assuring quality of machine translations based on large language models. Each aspect includes a series of steps for receiving data, translating the data, determining a confidence value, and revising the translation based on the confidence value.

Key Takeaways:

  • The proposed invention aims to improve the accuracy and reliability of machine translation by using large language models to revise translations that are below a certain confidence threshold.
  • The method, system, and computer program product for assuring quality of machine translations based on large language models include receiving data, translating the data, determining a confidence value, and revising the translation based on the confidence value.
  • The proposed invention uses a threshold value to determine when to revise the translation based on the confidence value.
  • The translation engine is trained based on comparing translation cases and known correct translations based on confidence values associated with a threshold value.
  • The proposed invention can be implemented using a computer system and a computer program product.

Statistics:

  • The patent application was filed on December 5, 2023.
  • The patent application was posted on June 5, 2025.
  • 20 claims are included in the patent application.
  • The proposed invention aims to improve the accuracy and reliability of machine translation by using large language models to revise translations that are below a certain confidence threshold.
  • The proposed invention can be implemented using a computer system and a computer program product.
  • The translation engine is trained based on comparing translation cases and known correct translations based on confidence values associated with a threshold value.

Sources:

  • "Translation Quality Assurance Based On Large Language Models" (patent application, December 5, 2023, posted June 5, 2025)
  • U.S. Patent Application Number 20250181335 (filed December 5, 2023 and posted June 5, 2025)
  • https://ppubs.uspto.gov/pubwebapp/external.html?q=(20250181335)&db=US-PGPUB&type=ids (patent URL)

Read more