AI-102 Question 417
Single answerYou are tasked with creating a composed document intelligence model in Azure AI to process invoices and receipts. The invoices have a fixed layout, while the receipts have variable structures. What is the best approach to implement the composed model in this scenario?
- A
Train a single custom model to handle both invoices and receipts.
- B
Use prebuilt models for both invoices and receipts without composing them.
- C
Create individual custom models for invoices and receipts, and then combine them into a composed model.
- D
Leverage the Azure Form Recognizer Layout model for both invoices and receipts without creating a composed model.
Show answer and explanation
Correct answer: C
Explanation
When creating a composed document intelligence model in Azure AI, the recommended practice is to train individual models for each document type (such as invoices and receipts) and then combine them into a composed model. This ensures accurate processing of documents with different structures and layouts, as the composed model can intelligently route each document to the appropriate sub-model.
- A. Incorrect.
This is not the best approach because invoices and receipts have different structures and would require individual models to achieve higher accuracy.
- B. Incorrect.
Using prebuilt models separately does not leverage the composed model feature, which is designed to handle scenarios with multiple document types.
- C. Correct.
This is the correct approach because a composed model allows you to combine individual models for different document types, ensuring better handling of both invoices and receipts.
- D. Incorrect.
The Layout model is not specialized for invoices or receipts. It provides basic layout extraction and is not suitable for building a composed model.