AI-102 Question 419
Select 3You are tasked with creating an Azure AI Search solution that processes scanned PDF documents containing forms and extracts key-value pairs as part of the search index. To achieve this, you decide to implement a custom Azure AI Search skill using Azure AI Document Intelligence (formerly known as Form Recognizer). Which of the following steps are required to successfully integrate this custom skill into your Azure Cognitive Search pipeline?
- A
Create an Azure AI Document Intelligence resource and configure a custom model to extract key-value pairs.
- B
Define a skillset in Azure Cognitive Search and include a custom web API skill that calls the Azure AI Document Intelligence resource.
- C
Add an indexer to your Cognitive Search pipeline to fetch and process data using the defined skillset.
- D
Deploy the custom skill as a managed endpoint in Azure Functions for real-time processing.
- E
Use the Azure AI Document Intelligence SDK to pre-process documents and manually push the results into the search index.
Show answer and explanation
Correct answers: A, B, C
Explanation
To integrate a document intelligence model as a custom Azure AI Search skill, you first need to create an Azure AI Document Intelligence resource to handle document processing. Then, define a skillset in Azure Cognitive Search that includes a custom web API skill to call the Document Intelligence resource. Finally, use an indexer to fetch and process data through the skillset, updating the search index. This approach ensures that your search solution is scalable and follows best practices in Azure Cognitive Search.
- A. Correct.
Correct. Azure AI Document Intelligence is required to extract structured data from documents, such as key-value pairs, which will be integrated into the Cognitive Search pipeline.
- B. Correct.
Correct. The custom web API skill enables the Cognitive Search pipeline to call the Azure AI Document Intelligence resource and process document data.
- C. Correct.
Correct. An indexer is necessary to connect the skillset to the data source and populate the search index with processed data.
- D. Incorrect.
Incorrect. While Azure Functions can be used for custom web API skills, it is not mandatory to deploy the custom skill as a managed endpoint in this scenario.
- E. Incorrect.
Incorrect. Pre-processing documents manually using the Azure AI Document Intelligence SDK and pushing the results directly to the search index bypasses the skillset and indexer configuration, which conflicts with the Azure Cognitive Search pipeline design.