AI-900 Practice Question: Describe features of computer vision workloads on Azure

Answer analysis

Option-by-option breakdown

For each option: why learners choose it and why it is or isn't the right answer here.

• ✗

  Automatically generating new document templates from existing forms

  Why it's wrong here

  Generating new document templates from existing forms is a document-design activity, not an extraction capability. A custom extraction model in Azure AI Document Intelligence is trained on labeled examples to learn where specific fields reside, then uses that model to extract those fields from future documents. It does not generate or modify templates; rather, it consumes the structure you already have and maps your business-specific field names to the data, leaving the original template untouched.

• ✓

  Training on your labeled documents to extract business-specific fields not covered by prebuilt models

  Why this is correct

  Custom extraction models are trained on your own labeled documents, where you tag the fields that matter to your business, so they can extract data that prebuilt models do not cover. Azure AI Document Intelligence lets you create a custom model by labeling a few sample documents with field names and positions, then the model learns those patterns and extracts those fields from new documents automatically. This directly matches the scenario of extracting business-specific fields from documents that don't fit standard prebuilt models, because the model is tailored to your exact forms.

• ✗

  Translating documents into multiple languages simultaneously

  Why it's wrong here

  Document translation is a native capability of Azure AI Translator, not Azure AI Document Intelligence's custom extraction. Custom extraction models are built to identify and retrieve structured data (field names and values) from forms and documents, not to change the language of the content. Even if translation were added as a preprocessing step, it would not produce the labeled field extraction the task requires; the core service here is custom extraction, which is language-agnostic and purely concerned with locating and extracting values.

• ✗

  Redacting sensitive information from documents automatically

  Why it's wrong here

  Redacting personally identifiable information (PII) is performed by separate tools such as Azure AI Language's PII detection or Document Intelligence's redaction features, not by custom extraction. Custom extraction models focus on pulling out specific field values — names of fields you define — from business forms, leaving all other content untouched and unaltered. They are not designed to identify and remove sensitive text across a document; that is a distinct, often post-processing, capability.

Go deeper

Related to this question