AI-102 Question 237
Single answerYou are developing a custom translation model using Azure AI Translator to translate legal documents from English to French. After uploading your custom training data and training the model, you notice that the translation quality is suboptimal for domain-specific terminology. What is the next step you should take to improve the model's performance?
- A
Add more general-purpose bilingual data to the training dataset.
- B
Use the Custom Translator to refine the model by providing additional in-domain training data.
- C
Publish the model as it is and rely on Translator’s built-in general translation engine to fill gaps.
- D
Enable the model's 'Auto-Improve' feature to automatically refine the translation quality over time.
Show answer and explanation
Correct answer: B
Explanation
To improve a custom translation model, it is essential to provide training data that is closely related to the domain you are targeting. In this scenario, providing additional in-domain training data through the Custom Translator ensures that the model can better handle the specialized terminology and context found in legal documents.
- A. Incorrect.
Adding more general-purpose bilingual data may dilute the domain-specific focus of the model, which is critical for translating legal terminology accurately.
- B. Correct.
Using the Custom Translator to refine the model by providing additional in-domain training data is the correct approach. This allows the model to better learn the specific terms and phrases used in legal documents.
- C. Incorrect.
Publishing the model without addressing the suboptimal quality would result in poor translation accuracy for the intended purpose. The general translation engine cannot reliably handle domain-specific terminology.
- D. Incorrect.
There is no 'Auto-Improve' feature in Azure AI Translator. Model improvement requires manual intervention, such as providing additional training or tuning data.