
AI-102 EXAM QUESTIONS 2026 MICROSOFT AZURE AI 102 CERTIFICATION COURSE PART-13
AI Summary
Welcome back to our AI 102 question series. Today, we’re diving into more questions to help you prepare for your exam. A big congratulations to Harit Singh for successfully clearing the AI 102 exam with a great score! He highlighted the importance of understanding questions related to content moderation and OpenAI, which we’ve covered extensively. Remember, the AI landscape is constantly changing, so always watch the latest videos and follow the playlist in sequence, as I often refer back to previous explanations.
We’re starting with question 111. You’re developing a method using the Azure AI Vision client library to perform OCR on images. The current code calls the `get_read_results` method before the OCR operation is complete, potentially causing errors. You need to prevent `get_read_results` from proceeding until the read operation is finished. Which two actions should you perform?
First, let’s examine the options. Removing the `operation_id` parameter would break the method’s ability to track and retrieve the OCR operation, leading to a total failure. So, this is incorrect.
Next, adding code to verify the `read_results.status` value is crucial. This allows the code to confirm if the read operation has succeeded before attempting to retrieve results, preventing errors from incomplete operations. This is a correct action.
Verifying the `read_operation.location` value might check URL validity but doesn’t guarantee the OCR process is complete, making it insufficient for our problem.
Finally, wrapping the call to `get_read_results` within a loop that includes a delay and checks the status is also an effective approach. This ensures the loop continues to check the operation’s status, likely looking for a ‘succeeded’ state, before proceeding. This is also a correct action. Therefore, the two correct actions are to verify the `read_results.status` value and to wrap the `get_read_results` call in a loop with a delay and status check.
Moving to question 112. You’re building a text-to-speech app using a custom neural voice and need to create an SSML file. The voice profile must express a calm tone and imitate a young adult female voice. How should you complete the SSML tags?
We’ve previously discussed SSML (Speech Synthesis Markup Language) in question 33, highlighting its power in controlling pronunciation, volume, speed, and emphasis. For this question, we’ll focus on customizing voice attributes.
The `voice` element in SSML allows you to specify the voice to be used. The `name` attribute within the `voice` tag is used to select a pre-built voice. Azure AI Speech offers a wide range of voices based on locale and accent. For instance, for English (India), you might find voices like `en-IN-PrabhatNeural` or `en-IN-AaryaNeural`.
The `effect` attribute within the `voice` element is used to control audio effect processors, optimizing speech output for specific scenarios like car speakers or telephone calls. Supported values include `EQ_CAR` for automotive environments and `EQ_TELECOM_HP8K` for telephone scenarios.
The `style` attribute controls the voice-specific speaking style, allowing you to express emotions like cheerfulness, empathy, or calmness. This is a required attribute when using the `mstts:express-as` element, especially for conveying emotions. If the `style` value is missing or invalid, the `mstts:express-as` element is ignored.
The `style_degree` attribute controls the intensity of the speaking style, with values ranging from 0.01 to 2, where 1 is the default.
The `role` attribute is used to imitate different ages or genders. In our case, to imitate a young adult female voice, we would use `role="young_adult_female"`.
To express a calm tone, we would use a `style` attribute, such as `style="calm"` or potentially `style="gentle"`. Considering the requirement for a calm tone, `style="gentle"` is a suitable choice.
Therefore, to complete the blanks: for the first blank, specifying the role, you should use `role="young_adult_female"`. For the second blank, specifying the style for the calm tone, you should use `style="gentle"`.
Next, question 113. Your app using Azure AI Speech service will be used in motor vehicles, and you need to optimize the synthesized voice output. Which SSML attribute should you configure?
Based on our previous discussion about the `effect` attribute, when dealing with automotive environments like motor vehicles, you would use `effect="EQ_CAR"` to optimize the auditory experience. Thus, the `effect` attribute of the `voice` element should be configured.
Question 114 asks about configuring a request for the Azure AI Vision API to identify if an image is clip art or a line drawing. How should you complete blank one of the request? The options are `GET`, `POST`, `PATCH`.
The `GET` method requests data without modification and is unsuitable for sending images for analysis. The `POST` method is used to send data to a server for processing, which is exactly what we need to do to analyze an image. The `PATCH` method is for partial modifications. Therefore, `POST` is the correct method.
Question 115 is a follow-up, asking what should come in blank two for the visual features. The options are `description`, `image_type`, `objects`, `tags`.
`description` generates a human-readable summary. `image_type` analyzes the kind of image (clip art, line drawing, etc.), which is precisely what we need. `objects` detects specific objects within the image, and `tags` generates descriptive keywords. Thus, `image_type` is the correct visual feature to configure.
Question 116 involves multiple data sources: on-premises SQL Server (finance), Cosmos DB (sales), Table Storage (logos), and SQL database (HR). You need to search all data using Azure AI Search. What should you do?
To enable efficient cloud-based indexing and search with Azure AI Search, data should ideally reside in Azure cloud storage services. Migrating the on-premises SQL Server data for finance to Azure Data Lake Storage is a key step. Azure AI Search integrates well with various Azure storage services, including Data Lake Storage, Cosmos DB, and Table Storage. Migrating the on-premises data to a cloud service like Azure Data Lake Storage will allow it to be indexed and searched alongside the other cloud-based data sources. Therefore, migrating the finance data to Azure Data Lake Storage is the recommended action.
Question 117 presents a Python method `create_resource` with parameters for resource name, kind, tier, and location. You need to deploy an Azure resource for sentiment analysis in the East US region. How should you call this method?
For sentiment analysis, the `kind` should be `text_analytics`. For production workloads requiring a balance of capabilities and cost, the `tier` should be `S0`. The `location` must be specified as `eastus` (lowercase, no spaces).
Question 118 asks for two responsible AI principles to guide monitoring for equitable results, regardless of user location or background, for a new sales system processing video and text from a public website. The six guiding principles are Fairness, Reliability and Safety, Privacy and Security, Inclusiveness, Transparency, and Accountability.
Fairness is directly related to ensuring equitable results without bias. Inclusiveness ensures the system is accessible and works for everyone. Therefore, Fairness and Inclusiveness are the two principles that provide guidance for meeting the monitoring requirements for equitable results.
Finally, question 119. You have 50,000 scanned documents for OCR and text analytics using cognitive search, and you need to minimize costs. What should you attach to the skill set?
You need a service that can perform both OCR and text analytics. While Computer Vision can do OCR, it doesn’t handle text analytics. Azure Cognitive Services provide these capabilities. Given the volume of 50,000 documents and the need to minimize costs, using a free limited enrichment cognitive services source would not be sufficient. The `S0` pricing tier for Azure Cognitive Services offers a good balance between cost-effectiveness and processing capabilities, making it the ideal choice for handling this volume of data. Therefore, you should attach an `S0` pricing tier Azure Cognitive Services source to the skill set.