Practice Test #2

Incorrect. Regenerating the primary admin key first would immediately invalidate the credential the app is currently using. That means the management portal could lose access until its configuration is updated to the secondary key, which does not minimize downtime. Although the final state would rotate both keys, the sequence is operationally wrong for a live app currently bound to the compromised primary key.

Incorrect. A query key cannot call index management endpoints in Azure Cognitive Search. Since the web app is a management portal for indexing, switching it to a query key would remove the permissions it needs to function. Regenerating the admin keys is useful, but this option breaks required administrative capabilities.

Incorrect. Query keys are only for read-only search requests and have no effect on admin-level index management access. Adding or deleting query keys does nothing to invalidate a compromised admin key that can modify indexes. This option also fails because the management portal cannot use a query key for administrative operations.

Incorrect. "CustomVision.Prediction" is used to host and run predictions for models you trained with Custom Vision (classification/object detection). It does not provide the prebuilt captioning capability that comes with Computer Vision/Image Analysis. Even though "F0" is free and "westus" is valid, the service kind is wrong for automatic caption generation.

Incorrect. "ComputerVision" is the correct service kind for captioning, but "S0" is the paid Standard tier, not the free tier. The question explicitly requires creating a free Azure resource, so using S0 violates the requirement even though it would work functionally for captioning.

Incorrect. This option is wrong on two dimensions: it uses "CustomVision.Prediction" (not the right service for prebuilt captioning) and it uses "S0" (paid tier) instead of the required free tier. Custom Vision prediction resources are intended for custom model inference, not generating captions from images.

Fairness focuses on ensuring the system does not produce biased or discriminatory results across different user groups (for example, across gender, age, or ethnicity). Simply notifying users that processing occurred does not address whether the model’s outputs are equitable or whether bias has been measured and mitigated. Fairness would require steps like representative training data, bias evaluation, and mitigation techniques. Therefore, a notification alone does not satisfy the fairness principle.

Inclusiveness is about designing AI systems that are accessible and usable by people with diverse abilities, languages, and backgrounds. A notification that data has been processed does not ensure the system supports accessibility needs (e.g., screen readers), multilingual users, or varied user contexts. Inclusiveness would involve UX and feature design choices that broaden participation and reduce barriers. Thus, inclusiveness is not the primary principle addressed by the notification.

Reliability and safety concern whether the system performs consistently, securely, and safely under expected and unexpected conditions. Notifying users about processing does not improve model robustness, error handling, security controls, or safety guardrails. Reliability and safety would be demonstrated through testing, monitoring, fallback behaviors, and protections against harmful outputs or failures. Therefore, this option does not match the scenario’s action of user notification.

Incorrect for this scenario. FileDTO objects are used when you create a KB from file content (for example, uploading documents). The question states the KBs are created from an array of URLs, so you don’t need to represent WebJob data as FileDTO. Even if files were involved, you would still need CreateKbDTO and CreateAsync; FileDTO alone is not sufficient.

Incorrect for the stated requirement. QnADTO is used when you supply explicit question/answer pairs (curated QnA) as the KB content. The question specifies creating KBs from an array of URLs, which are represented as URL sources in the create request. You could optionally add QnADTO entries, but it is not required to meet the URL-based creation requirement.

sentiment analyzes opinion polarity and returns scores/labels (positive, neutral, negative) and optionally sentence-level sentiment. While useful for understanding how customers feel, it does not provide the words or phrases needed to render a word cloud. You might combine sentiment with key phrases, but sentiment alone is insufficient.

languages detects the language of input text. This can be a helpful preprocessing step to route documents to the right models or provide language hints, improving downstream extraction quality. However, language detection does not extract keywords or phrases, so it cannot directly support building a word cloud.

entities/recognition/general performs named entity recognition (NER) to identify entities like organizations, people, locations, and other categories. This can help if your word cloud is specifically entity-focused, but typical review word clouds need descriptive key phrases (features, issues, attributes) that are better captured by key phrase extraction.

botId is not the standard required value for Bot Channels Registration deployment. In Bot Framework and Azure Bot Service, the key identity used by channels is the Azure AD application (Microsoft App) identity. Some tooling may display a bot resource ID, but it is not the credential pair required for authenticating requests to your bot endpoint.

tenantId identifies the Azure AD tenant where an app is registered. While it can matter for multi-tenant vs single-tenant app scenarios, Bot Channels Registration does not require tenantId to complete deployment. The bot’s runtime authentication with channels relies on the appId and a credential (secret/cert), not explicitly on tenantId.

objectId is the directory object identifier for the app registration’s service principal or application object. It is useful for Graph API operations and some Azure AD management tasks, but it is not required to configure Bot Channels Registration for a bot’s authentication. The bot framework expects the appId and a credential, not the objectId.

Incorrect. GetReadResultAsync requires the operation identifier to retrieve the OCR operation status and results. In the .NET SDK, this is typically passed as a Guid (hence Guid.Parse). Removing the parameter would not compile and does not address the asynchronous nature of the Read operation. The timing issue is solved by polling, not by changing how the ID is passed.

Incorrect. The object returned from ReadAsync primarily provides the Operation-Location header used to poll for results. The completion status is not reliably obtained from the initial response headers; instead, you must call GetReadResultAsync and inspect the returned ReadOperationResult.Status. Checking txtHeaders.Status is not the correct mechanism for determining when OCR processing has finished.

No is incorrect because moving to a higher tier is a recognized mitigation for throttling caused by insufficient capacity. Higher tiers generally offer increased performance characteristics and higher limits, which helps absorb increased query load. While scaling out replicas is often the most direct fix for query throughput, tier can be the limiting factor for how far you can scale and the baseline capacity available. Therefore, upgrading the tier can meet the goal of reducing throttling likelihood.

Cognitive Search is used for knowledge mining: indexing, enriching, and querying text and documents (including PDFs, Office files, and structured data). While it can store and search records, it does not provide native time-series anomaly detection algorithms for sensor telemetry. It would be useful for searching maintenance logs or manuals, not for detecting unusual numeric values in time series.

Form Recognizer (Azure AI Document Intelligence) extracts text, key-value pairs, tables, and structured fields from documents like invoices, receipts, and forms. It is not intended for analyzing numeric telemetry streams or detecting anomalies in time-series sensor data. It could help digitize maintenance reports, but it won’t identify outliers in minute-by-minute sensor readings.

Custom Vision is a computer vision service for training image classification or object detection models using labeled images. It is appropriate for visual inspection scenarios (e.g., detecting defects in photos), not for analyzing time-series sensor measurements. Since the problem is purely time-series anomaly detection across many sensors, Custom Vision does not fit the requirement.