AI-102 Question 214
Select 2You are designing a custom speech solution for a multilingual call center using Azure Speech Services. The call center requires the ability to transcribe calls in real-time in English and Spanish, and your solution must allow for custom vocabulary like product names and industry-specific terms. Which steps should you take to implement this solution?
- A
Create a Speech service resource in Azure and use the Speech-to-Text API directly for real-time transcription.
- B
Train and deploy a custom speech model using the Custom Speech portal to include domain-specific vocabulary.
- C
Use the Azure Translator Text API in combination with the Speech-to-Text API to handle multilingual transcription.
- D
Set up a language identification model to automatically detect the language of each call before sending it to the Speech-to-Text API.
- E
Integrate the Custom Commands feature of Azure Speech Services to enable command recognition for call center workflows.
Show answer and explanation
Correct answers: B, D
Explanation
To implement a custom speech solution for a multilingual call center, you need to ensure accurate transcription by including domain-specific vocabulary through a custom speech model. Additionally, since the call center handles multiple languages, a language identification model is necessary to route calls to the appropriate Speech-to-Text model. These steps ensure the solution meets the requirements for real-time, accurate multilingual transcription.
- A. Incorrect.
This option utilizes the basic Speech-to-Text API but does not account for custom vocabulary or language detection, making it insufficient for the scenario.
- B. Correct.
Correct. Training and deploying a custom speech model allows you to include domain-specific vocabulary, which is essential for accurate transcription in the call center.
- C. Incorrect.
The Azure Translator Text API is not necessary for this scenario, as the Speech-to-Text API already supports multilingual transcription when configured properly.
- D. Correct.
Correct. Setting up a language identification model ensures the correct language is detected and routed to the appropriate Speech-to-Text model for transcription.
- E. Incorrect.
The Custom Commands feature is primarily used for creating voice-powered applications with predefined commands and is not relevant to real-time multilingual transcription.