AI-102 Question 423
Select 3You are tasked with implementing a custom Azure AI Search skill that extracts structured data from invoices, such as invoice numbers, dates, and total amounts, to enhance search capabilities in a knowledge mining solution. Which steps should you include in your implementation pipeline?
- A
Integrate a pre-trained Azure Form Recognizer model as part of the skillset pipeline to extract key fields from the invoices.
- B
Configure the Azure AI Search skillset to output the extracted data as searchable fields in the Azure Cognitive Search index.
- C
Write a custom skill in Azure Functions to preprocess the invoices by converting them to a text-based format before passing them to the Azure Form Recognizer.
- D
Use Azure Blob Storage to store the invoices and configure Cognitive Search to directly index the data without using a skillset.
- E
Define a custom cognitive skill to translate the extracted data into multiple languages for multilingual search capabilities.
Show answer and explanation
Correct answers: A, B, C
Explanation
To implement a document intelligence model as a custom Azure AI Search skill, you need to use a pre-trained Azure Form Recognizer model to extract structured data from invoices. This data should then be configured as searchable fields in the Azure Cognitive Search index. Preprocessing the documents using a custom Azure Function skill might be necessary to ensure they are in a compatible format for extraction. The pipeline does not involve direct indexing without a skillset, and translating data into multiple languages is not a core requirement for this scenario.
- A. Correct.
Integrating a pre-trained Azure Form Recognizer model is essential for extracting structured data from invoices as part of the custom skillset pipeline.
- B. Correct.
Configuring the skillset to output the extracted data into searchable fields ensures the data becomes usable in the Azure Cognitive Search index.
- C. Correct.
Preprocessing the invoices using a custom Azure Function skill may be necessary if the invoices are not already in a format that the Azure Form Recognizer can process, such as converting images or PDFs to text.
- D. Incorrect.
Using Azure Blob Storage to directly index the data without a skillset is incorrect because the task requires structured data extraction, which necessitates a custom skillset pipeline.
- E. Incorrect.
While translating extracted data into multiple languages could be a feature, it is not a required step for implementing the document intelligence model in this scenario.