Your benefits with Adaptive Generative Translation
Adaptive Generative Translation – AGT. That almost sounds like… ChatGPT.
And it’s based on the same principle.
Instead of neural MT, a large language model (LLM) is used to generate translation suggestions.
Your data remains confidential
AGT works within the framework of the GDPR. Your data stays private. No MT engines are trained.
Efficient
AGT creates translation suggestions based on your content stored on the PRODOC memoQ server. Your terminology and common expressions are taken into account. This requires significantly less post-editing compared to neural machine translations such as Google Translator or DeepL.
Many languages
Many important languages are already supported in the evaluation phase: English, French, German, Spanish, Italian, Portuguese, Dutch, Russian, Romanian, Polish, Finnish, Danish, Swedish, Japanese, Norwegian, Chinese, Czech, Hungarian, Indonesian, Turkish, Croatian
How does AGT work?
AGT is an engine that is used for data exchange between the memoQ client at PRODOC and Microsoft Azure Open AI.
memoQ sends the text to be translated to AGT.
As additional information, memoQ also provides existing hits from the translation memory and fixed terms from the term database.
AGT uses this to generate an LLM prompt that is sent to Azure Open AI.
Azure Open AI provides a translation suggestion based on the prompt.
You can find more detailed information about AGT at https://agt.memoq.com/
Comparison AGT - neural MT
AGT
Based on a Large Language Model (LLM).
Uses probabilities, whose weighting is influenced by context information from the prompt, to predict the next best word.
The AI resides in the generation of the prompt and thus within memoQ.
Data remains confidential!
Neural MT
Based on a neural machine translation engine.
Uses probabilities determined from previous translations to predict the best translation of a sentence.
The AI resides with the respective machine translation provider.
Data travels to Google, DeepL & Co.