AI-102 Question 414
Select 3You are tasked with automating the processing of various types of documents, including invoices, receipts, and purchase orders, for your organization. Each document type has a distinct structure and set of fields to extract. You decide to use Azure Form Recognizer to build a composed document intelligence model. Which steps must you take to successfully create and deploy this model?
- A
Train individual custom models for each document type using labeled data
- B
Combine the trained models into a single composed model
- C
Deploy the composed model to an Azure App Service
- D
Test the composed model with sample documents to ensure accurate field extraction
- E
Create a single custom model for all document types without separating them
Show answer and explanation
Correct answers: A, B, D
Explanation
To create a composed document intelligence model in Azure Form Recognizer, you must first train individual custom models for each document type using labeled data. These models are then combined into a composed model, which allows the system to automatically classify and process documents based on their structure. After combining the models, testing the composed model ensures accurate extraction of fields from different document types. Deploying the model to Azure App Service is not necessary, and using a single model for all document types would lead to reduced accuracy.
- A. Correct.
This is correct because creating a composed model requires training individual custom models for each document type. Each model is specialized in extracting fields from a specific type of document.
- B. Correct.
This is correct because after training individual models, you must combine them into a composed model. This allows the system to automatically identify and process different document types based on their structure.
- C. Incorrect.
This is incorrect because deploying the composed model directly to an Azure App Service is not required. Instead, the composed model can be used through the Azure Form Recognizer service endpoint.
- D. Correct.
This is correct because testing the composed model with sample documents is a necessary step to ensure that it accurately extracts the required fields from each document type.
- E. Incorrect.
This is incorrect because using a single custom model for all document types without separating them will result in poor accuracy. Each document type should have its own model to account for structural differences.