AI-102 Question 239
Select 3You are tasked with building a custom translation model using Azure Translator for an organization that requires highly accurate translations of legal documents. After creating and training the custom model, you notice the accuracy is still not meeting the requirements. Which of the following actions should you take to enhance the model's performance before publishing?
- A
Add more domain-specific training data to the model.
- B
Use the pre-trained baseline model without any custom training.
- C
Provide human feedback on incorrect translations to improve the model.
- D
Fine-tune the model using the custom translation evaluation tool.
- E
Increase the number of languages supported by the model to improve accuracy.
Show answer and explanation
Correct answers: A, C, D
Explanation
To enhance the performance of a custom translation model, focusing on domain-specific training data, incorporating human feedback, and fine-tuning the model are essential steps. These actions directly target the improvement of translation accuracy for specialized content like legal documents. Simply using the baseline model or increasing the number of supported languages does not address the specific issue of translation quality for the required domain.
- A. Correct.
Adding more domain-specific training data helps the model better understand the context and terminology of legal documents, improving translation accuracy.
- B. Incorrect.
Using the pre-trained baseline model without any custom training would not address the specific needs of translating legal documents, so it would not improve performance.
- C. Correct.
Providing human feedback on incorrect translations is a critical step in improving the model's accuracy, as it allows the system to learn from its mistakes.
- D. Correct.
Fine-tuning the model using the custom translation evaluation tool helps refine the model's performance based on evaluation metrics and further optimizes its accuracy.
- E. Incorrect.
Increasing the number of languages supported does not directly improve accuracy for a specific language or domain. It only broadens the model's linguistic capabilities.