A contact center needs live call transcripts translated for agents, and product names keep being misheard. What do you implement?
Why B is the answer
A speech pipeline with a custom speech model and translation
Live audio needs speech to text, custom speech models handle domain terms, and translation serves the agents. A nightly job cannot help during the call.
In detail
- A custom speech model handles product names.
- Translation is added to a live pipeline.
- Train custom speech with your product vocabulary.
Use case
An illustrative scenario, not a real incident.
A call center needs live translated transcripts, and product names are misheard. A custom speech model fixes the names, and translation runs in the same pipeline.
See it in different ways
flowchart LR A["Live audio"] --> B["Custom speech model"] B --> C["Transcript"] C --> D["Translation"] D --> E["Agent screen"]
flowchart LR S0["Call audio"] --> S1["Custom speech model"] --> S2["Transcript"] --> S3["Translation"]
Key point: Custom speech fixes specialist words.
Why the other options are wrong
- A. A nightly text analysis job over stored transcripts
- A nightly job cannot help on a live call.
- C. A retrieval index built directly over the raw call audio
- An index does not transcribe audio.
- D. An image generation flow that turns each call into a card
- Image generation is unrelated.