You want topics and a summary from support emails, in a format you define. What is a flexible option?
Why B is the answer
A language model with a prompt and a schema
A language model with a prompt and a schema can extract entities, topics and summaries in the format you define.
In detail
- A language model with a prompt and a schema can return exactly what you define.
- Validate the output against the schema.
Use case
An illustrative scenario, not a real incident.
A support team wants topics and a summary from emails in its own JSON format. A language model with a prompt and schema fits exactly.
See it in different ways
flowchart TD
A{"Text task"} -->|"Find personal data"| B["PII detection in Azure Language"]
A -->|"Change language"| C["Azure Translator"]
A -->|"Flexible analysis"| D["Language model with a prompt"]mindmap
root((Language tasks))
PII
Azure Language
Translation
Azure Translator
Custom format
Language model with schema
Safety
Groundedness detection
Protected materialflowchart LR S0["Email"] --> S1["Prompt with schema"] --> S2["Language model"] --> S3["JSON"]
Key point: A prompt and schema give flexible extraction.
Why the other options are wrong
- A. An embedding model with no prompt or schema
- Embeddings return vectors, not topics or summaries.
- C. A virtual network rule
- A network rule does not process text.
- D. A TPM quota increase
- Quota does not process text.