AI-102 Question 390
Single answerYour organization needs to process legal documents in bulk and extract specific fields such as 'Client Name', 'Contract Date', and 'Contract Amount'. You are tasked with building a solution using Azure AI Document Intelligence (formerly Azure Form Recognizer). The documents are semi-structured and vary slightly in format. Which approach should you take to implement the solution?
- A
Use the prebuilt layout model to extract the required fields.
- B
Train a custom model using labeled data that includes examples of the specific fields to extract.
- C
Use the prebuilt receipt model since it is optimized for extracting structured data from receipts.
- D
Manually parse the documents using OCR output from the Azure Computer Vision API.
Show answer and explanation
Correct answer: B
Explanation
The correct approach for extracting specific fields like 'Client Name', 'Contract Date', and 'Contract Amount' from semi-structured legal documents is to train a custom model using Azure AI Document Intelligence. This allows you to define and train the model to recognize and extract the required fields, even when the document formats vary slightly. While prebuilt models and OCR tools have their use cases, they are not suitable for custom field extraction in this scenario.
- A. Incorrect.
The prebuilt layout model is designed for extracting the general structure of documents, such as text, tables, and layout elements, but it does not specifically extract custom fields like 'Client Name', 'Contract Date', or 'Contract Amount'.
- B. Correct.
Training a custom model using Azure AI Document Intelligence allows you to define and extract specific fields from semi-structured or unstructured documents by providing labeled training data. This is the correct approach for this scenario.
- C. Incorrect.
The prebuilt receipt model is specifically tailored for extracting information from receipts, such as totals, tax, and merchant details. It is not suitable for processing legal documents or extracting fields like 'Contract Date'.
- D. Incorrect.
The Azure Computer Vision API provides OCR capabilities but does not offer the structured data extraction or field-specific training capabilities required for this task. Parsing raw OCR output manually would be complex and error-prone.