AI-102 Question 418
Single answerYou are designing a composed document intelligence model in Azure AI to process invoices from different vendors. Each vendor uses a distinct invoice format. How should you set up the composed model to ensure accurate extraction of data from all invoice formats?
- A
Combine multiple prebuilt models into a single composed model.
- B
Train a custom model for each vendor's invoice format, then compose them into a single model.
- C
Use a single custom model trained with sample data from all vendors' invoice formats.
- D
Deploy a prebuilt receipt model to handle generic invoice data extraction.
Show answer and explanation
Correct answer: B
Explanation
When working with multiple distinct document formats, such as invoices from different vendors, the best practice is to train a custom model for each format. These custom models can then be composed into a single composed model, enabling the system to automatically select the most appropriate model for a given document. This approach ensures high accuracy and adaptability for diverse document types.
- A. Incorrect.
Prebuilt models are designed for general use cases and cannot be combined in a composed model for specific formats. This approach would not work for vendor-specific invoice processing.
- B. Correct.
This is the correct approach. By training a custom model for each vendor's invoice format, you can ensure accurate data extraction tailored to the specific layout and data points. Composing these models allows the system to automatically select the best-fit model for each document.
- C. Incorrect.
Using a single custom model trained with data from all vendors may lead to inaccuracies, as it would struggle to generalize over multiple distinct formats.
- D. Incorrect.
Prebuilt receipt models are not designed for invoice-specific extraction and may not capture all necessary data fields accurately. Additionally, they cannot be used in a composed model for custom vendor-specific formats.