AI-102 Question 408
Select 3You are tasked with extracting structured data from scanned invoices using Azure AI services. Your organization has unique invoice formats that are not supported by the prebuilt document models in Azure Form Recognizer. What steps should you take to implement a custom solution that meets this requirement?
- A
Label a training dataset with key-value pairs and tables in your invoices using the Form Recognizer Sample Labeling tool.
- B
Use the prebuilt invoice model in Azure Form Recognizer and apply additional post-processing logic to handle unsupported formats.
- C
Train a custom document intelligence model by uploading labeled training data to the Form Recognizer Studio and defining a custom model.
- D
Deploy the trained model and integrate it into your application using the Form Recognizer API endpoints.
- E
Enable Azure Cognitive Search and use it to extract data from the scanned invoices instead of using Form Recognizer.
Show answer and explanation
Correct answers: A, C, D
Explanation
To extract data from unique invoice formats, the prebuilt models in Azure Form Recognizer are insufficient, requiring you to train a custom model. This involves labeling a representative dataset, training the model in Form Recognizer Studio, and then deploying the model for use via the API. These steps ensure the solution is tailored to your organizational needs and provides accurate data extraction.
- A. Correct.
Correct. Labeling your training dataset is a necessary step to train a custom document model tailored to your unique invoice formats.
- B. Incorrect.
Incorrect. The prebuilt invoice model is not suitable in this case because your organization’s invoice formats are not supported, and additional post-processing logic cannot fully address this limitation.
- C. Correct.
Correct. After preparing labeled training data, you must train a custom model in Form Recognizer Studio to extract structured data specific to your invoice formats.
- D. Correct.
Correct. Once the custom model is trained, deploying it and integrating it into your application using Form Recognizer APIs is a required step to operationalize the solution.
- E. Incorrect.
Incorrect. Azure Cognitive Search is not designed for extracting structured data from scanned documents; it is primarily used for indexing and searching text.