AI-102 Question 211
Select 3You are designing a custom speech solution for a retail company. The company wants to transcribe customer service call recordings and ensure the transcription is tailored to their specific product names, industry jargon, and brand-specific terminology. Which steps should you take to implement this solution using Azure Speech Service?
- A
Train a Custom Speech model using audio data and a custom language model.
- B
Use the standard Microsoft Speech-to-Text API without additional customization.
- C
Create a custom pronunciation file to ensure correct pronunciation of brand-specific words.
- D
Upload a list of product names and industry-specific terminology as part of a Custom Speech project.
- E
Enable real-time transcription without any pre-training to prioritize speed over accuracy.
Show answer and explanation
Correct answers: A, C, D
Explanation
To implement a custom speech solution tailored to the company’s needs, you should train a Custom Speech model using relevant audio data and a custom language model. Additionally, creating a custom pronunciation file and uploading a list of product names and industry-specific terms ensures an accurate transcription of brand-specific language. Using the standard Speech-to-Text API or opting for real-time transcription without customization would not meet the company's requirements for accuracy and industry relevance.
- A. Correct.
Correct: Training a Custom Speech model with audio data and a custom language model ensures the service is optimized for the company's industry-specific needs and terminology.
- B. Incorrect.
Incorrect: The standard Microsoft Speech-to-Text API does not account for custom terminology or brand-specific language without customization.
- C. Correct.
Correct: A custom pronunciation file is essential for accurately transcribing unique or uncommon words related to the company’s products and brand.
- D. Correct.
Correct: Uploading a list of product names and industry-specific terminology helps the Custom Speech model recognize and correctly transcribe these terms.
- E. Incorrect.
Incorrect: While real-time transcription is possible, it does not allow for the integration of customizations such as specific terminology or pronunciation, leading to lower accuracy.