AI-102 Question 221
Single answerYou are tasked with implementing a keyword recognition feature in your Azure-based customer support system. The system should detect specific keywords such as 'help', 'agent', and 'cancel' from user voice inputs and trigger appropriate workflows. You decide to use Azure Speech Services to achieve this. What is the MOST appropriate way to implement keyword recognition in this scenario?
- A
Use the Azure Speech-to-Text service and write custom logic to analyze the transcribed text for specific keywords.
- B
Use the Azure Custom Keyword Recognizer feature in Speech SDK to detect predefined keywords directly in audio input.
- C
Use Azure Form Recognizer to process the audio input and extract keywords.
- D
Implement an Azure Bot Service with QnA Maker to identify the keywords from transcribed text.
Show answer and explanation
Correct answer: B
Explanation
The Azure Custom Keyword Recognizer feature in Speech SDK is specifically designed to detect predefined keywords directly from audio input. It provides a streamlined and efficient way to implement keyword recognition without requiring additional post-processing steps. This makes it the most appropriate solution for the described scenario.
- A. Incorrect.
While the Speech-to-Text service can be used to transcribe audio to text, it does not natively support keyword recognition. Writing custom logic would require additional coding and is less efficient than using a dedicated keyword recognition feature.
- B. Correct.
The Azure Custom Keyword Recognizer, available in the Speech SDK, is specifically designed for detecting predefined keywords in audio input, making it the most appropriate solution for this scenario.
- C. Incorrect.
Azure Form Recognizer is used for extracting information from structured and semi-structured documents, such as invoices or forms. It is not suitable for processing audio inputs or performing keyword recognition.
- D. Incorrect.
Azure Bot Service and QnA Maker are designed for building conversational AI and question-answering systems. While they could process text, they are not designed for direct keyword recognition from audio input.