Free Microsoft AI-102 Practice Test Questions MCQs

Explanation
Azure AI Immersive Reader is specifically designed to improve reading comprehension for users with learning differences (e.g., dyslexia) and reduced comprehension. It provides features like text sizing, spacing, syllable breakdown, part‑of‑speech highlighting, line focus, and picture dictionary, making it the correct choice for this scenario.

Correct Option

C. Azure AI Immersive Reader
Immersive Reader is an Azure Cognitive Service that helps users of all ages and abilities read and comprehend text more easily. It includes customizable display options, grammar support, and read‑aloud functionality, directly addressing the requirement.

Why Other Options Are Incorrect

A. Azure AI Translator –
Translates text between languages; does not provide reading optimization for comprehension or dyslexia.

B. Azure AI Document Intelligence –
Extracts structured data from documents; not for reading experience optimization.

D. Azure AI Language –
Performs text analytics (sentiment, key phrases, NER); does not offer immersive reading features.

Reference
Microsoft Learn: What is Azure AI Immersive Reader? – Designed for users with dyslexia and other learning differences.

Explanation
To convert customer calls (spoken audio) into text queries, you need speech recognition. The SpeechRecognizer class (from Azure AI Speech) converts speech to text. To provide a spoken response (text‑to‑speech), you need SpeechSynthesizer. These are the standard, low‑effort APIs for this scenario.

Correct Options

First requirement (convert customer calls into text queries): SpeechRecognizer
SpeechRecognizer performs speech‑to‑text conversion, turning the customer's spoken words into text that can be used to query the database.

Second requirement (provide customers with the order details as spoken response): SpeechSynthesizer
SpeechSynthesizer performs text‑to‑speech conversion, generating a spoken response from the order details retrieved from the database.

Why Other Options Are Incorrect

TranslationRecognizer – Used for speech translation, not simple speech‑to‑text or text‑to‑speech.

VoiceProfileClient – Used for speaker recognition/identification, not for converting speech or synthesizing responses.

Reference
Microsoft Learn: Azure AI Speech – SpeechRecognizer – Converts audio to text.

Explanation
To upload datasets and fine‑tune models in Azure OpenAI, User1 needs permissions beyond just calling the API (which is the Cognitive Services OpenAI User role). The Cognitive Services OpenAI Contributor role allows creating, training, and deploying models, as well as uploading data, while following least privilege.

Correct Option

D. Cognitive Services OpenAI Contributor
This built‑in role grants permissions to manage Azure OpenAI resources, including fine‑tuning models and uploading datasets. It is more permissive than Cognitive Services OpenAI User (which only allows inference) but less permissive than full Contributor, adhering to least privilege.

Why Other Options Are Incorrect

A. Cognitive Services Contributor –
Grants access to all Cognitive Services resources (not just OpenAI) and is broader than necessary.

B. Contributor –
Full management access to all Azure resources; violates least privilege.

C. Cognitive Services OpenAI User –
Allows only inference (calling deployed models), not fine‑tuning or dataset upload.

Reference
Microsoft Learn: Azure OpenAI RBAC roles – Cognitive Services OpenAI Contributor allows fine‑tuning and model management.

Explanation
To detect sexually explicit images with minimal development effort, you should use Azure AI Content Safety Studio, which provides pre‑built image moderation for hate, sexual, violence, and self‑harm content. You can test and integrate the API without training custom models.

Correct Option

C. Azure AI Content Safety Studio
Content Safety Studio offers a dedicated image moderation feature that analyzes images for sexual content and returns severity levels. It is designed specifically for this use case, requires no training, and minimizes development effort.

Why Other Options Are Incorrect

A. Visual Studio –
An IDE for coding, not a service for content moderation.

B. Vision Studio in Azure AI Vision –
Vision Studio includes adult/racy detection but is part of Computer Vision. Content Safety is the dedicated service for this scenario.

D. Azure AI Studio –
A general platform for building AI solutions; not a specific tool for image moderation.

Reference
Microsoft Learn: Azure AI Content Safety – Image moderation – Detects sexual content with severity levels.

Explanation
To identify an Azure AI services account after creation, you need to retrieve its details (endpoint, keys, location, etc.). The az cognitiveservices account show command returns the full account information, including the resource ID, endpoint, and properties. This is the correct command with minimal effort.

Correct Option

D. az cognitiveservices account show
This command displays details of a specific Cognitive Services account, such as its endpoint, location, SKU, and provisioning state. It is the standard CLI command to identify and verify an account after creation.

Why Other Options Are Incorrect

A. az resource link – Used for creating links between resources, not for identifying an account.

B. az account list – Lists Azure subscriptions, not Cognitive Services accounts.

C. az cognitiveservices account network-rule – Manages network access rules (firewall, virtual networks), not for identifying the account.

Reference
Microsoft Learn: az cognitiveservices account show – Retrieves account properties including endpoint, keys, and SKU.

Explanation
For handwritten survey responses with inconsistent layouts, a custom model (template or neural) is required because pre‑built models expect consistent form structures. In Document Intelligence Studio, when creating a project, you must first upload the training images to Azure Blob Storage (or a similar Azure Storage container), then point the project to that container.

Correct Options

Upload images to: Azure Blob Storage
Document Intelligence projects require training data to be stored in an Azure Blob Storage container. You upload the scanned images to a blob container, then connect the container to the project.

Type of model: Custom model (or Custom neural model)
Since the surveys have no consistent layout, a custom neural model is the best choice. Neural models handle unstructured, varied layouts and handwritten text better than template models. Pre‑built models (e.g., prebuilt-layout, prebuilt-document) are not designed for extracting custom data fields without training.

Why Other Options Are Incorrect

Azure File Share / OneDrive / Local upload – Document Intelligence Studio does not directly accept local uploads for training; it requires Azure Blob Storage.

Pre‑built model – Pre‑built models (e.g., prebuilt-invoice, prebuilt-receipt) are for specific document types with consistent layouts, not for varied, hand‑written surveys.

Template model – Works for documents with fixed layouts, but neural models are better for inconsistent layouts.

Reference
Microsoft Learn: Document Intelligence – Custom models – Neural models handle unstructured, handwritten documents.

Explanation
The code processes the result of a speech translation recognition. The TranslationRecognitionResult object has a Reason property to determine the outcome. For a successful translation, the Reason should be TranslatedSpeech. The translated text is stored in the Text property, and the Translations dictionary contains the translated outputs per target language.

Correct Options

First blank (after translationRecognitionResult.): Reason
The Reason property (of type ResultReason) indicates whether the recognition was successful (TranslatedSpeech), failed (NoMatch), or is ongoing (RecognizingSpeech). This is the correct property to switch on.

Second blank (case): TranslatedSpeech
When the recognition successfully translates the speech, ResultReason.TranslatedSpeech is returned. This case block handles the successful translation.

Why Other Options Are Incorrect

text – Not a property that indicates the result reason.

translations – Contains the translation outputs but does not indicate success/failure.

RecognizingSpeech – Indicates an interim result (partial recognition), not the final result.

NoMatch – Used in the fallback case, not the success case.

Reference
Microsoft Learn: Speech Translation – Recognition results – TranslationRecognitionResult.Reason with ResultReason.TranslatedSpeech for success.

Explanation
Content Understanding projects use templates to define the extraction schema. For analyzing sentiment in video files, the Media asset management template is designed to process video assets and extract metadata including sentiment, key moments, and descriptive attributes. This minimizes development effort compared to building a custom schema from scratch.

Correct Option

B. Media asset management
This template is built for processing video files (e.g., libraries, archives, media assets). It includes pre-configured capabilities for sentiment analysis, scene detection, and content summarization, reducing the need for custom configuration.

Why Other Options Are Incorrect

A. Video shot analysis –
Focuses on technical shot detection (cuts, transitions, camera movement), not sentiment analysis.

C. Advertising –
Optimized for ad creative analysis (product placement, brand logos, call-to-action), not general video sentiment.

Reference
Microsoft Learn: Azure AI Content Understanding – Templates – Media asset management template supports sentiment analysis for video libraries.

Explanation
Entity linking identifies named entities in text and links them to a knowledge base (Wikipedia). This provides additional context by returning a Wikipedia URL for each linked entity. It is the correct feature for adding references to supporting Wikipedia articles.

Correct Option

A. entity linking
Entity linking disambiguates entities (e.g., "Paris" as city vs person) and returns a url property pointing to a Wikipedia article. This gives users direct access to supporting articles, fulfilling the requirement.

Why Other Options Are Incorrect

B. custom NER – Extracts custom entities you define but does not link to Wikipedia. Requires training and does not provide external references.

C. Azure AI Content Safety – Detects objectionable content; does not link to Wikipedia.

D. key phrase extraction – Returns important topics but does not link to external articles.

Reference
Microsoft Learn: Entity linking in Azure AI Language – Returns Wikipedia URLs for recognized entities.

Explanation
To generate fictional stories, you need an Azure OpenAI resource. To use a customer‑managed key (CMK), you must specify the key vault URI, key name, and key version using the --api-properties parameter (or --encryption in newer CLI versions), and also enable a system‑assigned or user‑assigned identity with --assign-identity.

Correct Options

First blank (--kind): OpenAI
The resource kind for Azure OpenAI is OpenAI. This creates a resource capable of generating text responses (fictional stories).

Second blank (--assign-identity): --assign-identity
Required to assign a system‑assigned managed identity to the resource. The identity is used to authenticate to Azure Key Vault for accessing the customer‑managed key.

Third blank (key properties):
--api-properties (or --encryption) with the key vault URI, key name, and key version.

From the answer area, the correct structure is:

text
--api-properties keyVaultUri="https://issue23056kv.vault.azure.net/" keyName="KeyName" keyVersion="secretVersion"

Why Other Options Are Incorrect

Omission of --assign-identity – Without a managed identity, the resource cannot access Key Vault.

Incorrect key properties format – Missing keyName or keyVersion would cause CMK setup to fail.

Using --keyVaultUri alone – The full syntax requires keyVaultUri, keyName, and keyVersion.

Reference
Microsoft Learn: Create Cognitive Services resource with customer-managed key – Use --assign-identity and --api-properties keyVaultUri=... keyName=... keyVersion=....