AI-102 Question 212
Select 3You are designing a custom speech solution for a retail company that wants to transcribe customer service calls in real-time. The transcriptions must recognize industry-specific terms such as product names and abbreviations, and the solution must support multiple languages. Which steps should you take to implement the solution using Azure Cognitive Services?
- A
Train a custom acoustic model using Azure Speech Studio.
- B
Create a custom language model and upload industry-specific terms using the Speech-to-Text Customization feature.
- C
Enable real-time transcription using the Azure Speech SDK.
- D
Use the Translation feature in Azure Cognitive Services to support multiple languages.
- E
Deploy the custom models to an Azure Speech endpoint for real-time processing.
Show answer and explanation
Correct answers: B, C, E
Explanation
To implement a custom speech solution with Azure, you need to create and deploy a custom language model to recognize industry-specific terms. The Azure Speech SDK enables real-time transcription, and deploying the models to an endpoint allows the solution to process calls in real-time. Custom acoustic model training is not supported, and while the Translation feature supports multiple languages, it does not fulfill the requirement for real-time transcription.
- A. Incorrect.
Azure does not currently support custom acoustic model training. The Speech-to-Text Customization feature allows for customizing language, not acoustic models.
- B. Correct.
Creating a custom language model and uploading industry-specific terms ensures that the speech service can accurately recognize domain-specific vocabulary.
- C. Correct.
The Azure Speech SDK enables real-time transcription, which is a requirement for the solution.
- D. Incorrect.
While the Translation feature supports multiple languages, it is not directly related to real-time transcription or customization of speech solutions.
- E. Correct.
Deploying the custom models to an Azure Speech endpoint is the final step required to enable real-time transcription with the customizations applied.