AI-102 Question 406
Select 3You are tasked with building a custom document intelligence model using Azure AI to extract key-value pairs from a set of invoices. The invoices have various formats and contain tables, dates, and customer details. You decide to use the Azure Form Recognizer service. What steps should you take to implement the custom model effectively?
- A
Use the prebuilt invoice model in Azure Form Recognizer without any customization.
- B
Label a diverse set of sample invoices to train the custom model using the Form Recognizer Studio.
- C
Train a custom model by uploading labeled data and validating the model's performance.
- D
Use Azure Cognitive Services Vision API instead of Form Recognizer for document processing.
- E
Test the custom model with a new set of invoices to ensure it generalizes well to unseen data.
Show answer and explanation
Correct answers: B, C, E
Explanation
To implement a custom document intelligence model using Azure Form Recognizer, you need to label a representative set of sample documents, train the model on this labeled data, and validate its performance to ensure it meets your requirements. Additionally, testing the model on unseen data is essential to confirm it generalizes well to real-world scenarios. Using the prebuilt model or a different service like the Vision API would not provide the level of customization and specialization required for extracting specific fields from your invoices.
- A. Incorrect.
Using the prebuilt invoice model might be sufficient for simple use cases, but it won't allow you to tailor the model to the specific key-value pairs and formats in your invoices.
- B. Correct.
Labeling a diverse set of sample invoices is crucial for training a custom model, as it helps the AI learn the specific patterns and fields in your documents.
- C. Correct.
Training a custom model with labeled data and validating its performance is a necessary step for building a robust model tailored to your invoices.
- D. Incorrect.
The Vision API is designed for image analysis and does not specialize in document processing or extracting structured data like Form Recognizer does.
- E. Correct.
Testing the custom model with new data ensures the model performs well on unseen documents and generalizes effectively.