Your app must answer spoken questions out loud. Which capabilities do you need?
Why A is the answer
Speech to text and text to speech
You need to turn the speech into text for the model and turn its answer back into audio.
In detail
- Speech to text hears the question.
- Text to speech says the answer.
- A language model can sit between the two.
Use case
An illustrative scenario, not a real incident.
A hotel kiosk hears a guest's question and answers aloud. Speech to text handles the listening, and text to speech handles the reply.
See it in different ways
flowchart LR A["Spoken question"] --> B["Speech to text"] B --> C["Model answers"] C --> D["Text to speech"] D --> E["Spoken answer"]
sequenceDiagram participant U as User participant S as Azure Speech participant M as Model U->>S: Spoken question S->>M: Text M-->>S: Text reply S-->>U: Spoken reply
sequenceDiagram participant P0 as Guest participant P1 as Speech to text participant P2 as Model participant P3 as Text to speech P0->>P1: Spoken question P1->>P2: Text P2->>P3: Text reply P3-->>P0: Spoken reply
Key point: Hearing and speaking are two capabilities.
Why the other options are wrong
- B. Image generation and translation
- Neither of these handles audio.
- C. Document extraction and summarization
- These work on documents.
- D. Entity detection and key phrase extraction
- These work on text.