View Full Microsoft AI-901 Exam Dumps and Practice Test Dumps

Question 321. Which Azure service can be used to manage and deploy machine learning models and workflows?

1. Azure Machine Learning
2. Azure DNS
3. Azure Firewall
4. Azure CDN

Correct Answer: 1. Azure Machine Learning

Explanation:

Azure Machine Learning is a cloud service designed to support the development, training, deployment, and management of machine learning models. It provides capabilities for working with datasets, training jobs, compute resources, model assets, endpoints, and monitoring. Teams can use it throughout the machine learning lifecycle rather than relying on separate unmanaged tools for each stage. Azure Machine Learning can support both experimentation and production deployment scenarios. The service also provides capabilities that help organizations organize machine learning resources and manage models more consistently. Developers should still select the specific features and deployment patterns that match their application’s requirements.

Question 322. Which component provides computing resources for running machine learning training jobs?

1. Compute resource
2. Prompt template
3. Search index
4. Language detector

Correct Answer: 1. Compute resource

Explanation:

A compute resource provides the processing capacity required to run machine learning workloads. Training jobs can require significant CPU or GPU resources depending on the size of the dataset, model architecture, and complexity of the training process. Azure Machine Learning supports different compute options so teams can select resources appropriate for their workloads. Using managed compute can help organizations scale training resources when needed rather than maintaining all processing infrastructure themselves. The choice of compute should consider performance, cost, workload requirements, and availability. Efficient compute selection is important because unnecessarily large resources can increase costs without providing meaningful benefits.

Question 323. What is the main purpose of a machine learning pipeline?

1. To organize multiple steps in a repeatable workflow
2. To replace all training data
3. To store passwords
4. To translate speech

Correct Answer: 1. To organize multiple steps in a repeatable workflow

Explanation:

A machine learning pipeline organizes related steps into a repeatable workflow. A pipeline might include data preparation, feature engineering, model training, evaluation, and other processing activities. Automating these steps can improve consistency and make it easier to rerun a workflow when data or model configurations change. Pipelines are particularly useful when teams need repeatable machine learning processes rather than manually performing each stage. They can also support operational workflows in which data is processed regularly and models are retrained or evaluated according to defined conditions. The exact pipeline structure should reflect the requirements of the machine learning solution.

Question 324. Which Azure Machine Learning feature helps store and manage trained models for later use?

1. Model registry
2. Speech synthesizer
3. DNS zone
4. Network interface

Correct Answer: 1. Model registry

Explanation:

A model registry provides a managed location for registering and organizing trained machine learning models. It can help teams track model versions and identify which model artifacts are available for deployment or further evaluation. This becomes particularly valuable when multiple versions of a model are created during experimentation and production development. Version management can also support controlled deployment and rollback processes. A registry does not automatically determine whether a model is accurate or suitable for production. Each model should still be evaluated against appropriate requirements, but maintaining organized model records makes the overall machine learning lifecycle easier to manage.

Question 325. What is the purpose of model deployment?

1. To make a trained model available for inference
2. To delete training data
3. To create a new programming language
4. To translate documents

Correct Answer: 1. To make a trained model available for inference

Explanation:

Model deployment makes a trained machine learning model available so that applications or users can obtain predictions from it. Depending on the scenario, the model may be deployed through an online endpoint for real-time requests or a batch endpoint for processing larger workloads asynchronously. Deployment is more than simply placing a model file somewhere because teams also need to consider authentication, scaling, monitoring, dependencies, versioning, and operational requirements. A model should generally be evaluated before deployment and monitored afterward. This helps ensure that the deployed solution continues to meet its intended performance and reliability requirements.

Question 326. Which Azure AI Language capability can identify important words or phrases in a text?

1. Key phrase extraction
2. Object detection
3. Image segmentation
4. Speech synthesis

Correct Answer: 1. Key phrase extraction

Explanation:

Key phrase extraction identifies important concepts or phrases within unstructured text. It can help applications quickly determine the main topics discussed in documents, customer feedback, reviews, or other textual content. For example, a support message might contain key phrases related to account access, payment problems, or product issues. This information can be used for categorization, search, analytics, or downstream processing. Key phrase extraction is different from summarization because it focuses on identifying significant terms or phrases rather than generating a condensed version of the entire text. Applications should still evaluate extracted phrases against their specific domain and language requirements.

Question 327. Which Azure AI Language capability identifies entities such as people, organizations, and locations in text?

1. Named Entity Recognition
2. Image classification
3. Speech-to-text
4. Regression

Correct Answer: 1. Named Entity Recognition

Explanation:

Named Entity Recognition, or NER, identifies and categorizes entities mentioned in text. Common entity types can include people, organizations, locations, dates, and other recognized categories depending on the service and language. NER can help applications structure information from unstructured text, improve search, support analytics, and extract useful information from documents. For example, an organization could use NER to identify company names and locations in customer communications. The results should still be reviewed for the application’s specific domain because specialized terminology, unusual names, and ambiguous language can affect recognition quality.

Question 328. Which Azure AI Language capability can determine the language used in a text document?

1. Language detection
2. Object tracking
3. Regression
4. Image segmentation

Correct Answer: 1. Language detection

Explanation:

Language detection identifies the language or languages represented in a text input. This capability can be useful in applications that receive multilingual content and need to select an appropriate downstream process. For example, an application might detect whether a customer message is written in English, French, Spanish, or another supported language before routing it to translation or language-specific analysis. Language detection can be particularly useful in global applications where users are not required to select a language manually. Developers should test the service with realistic text, especially when inputs are short, mixed-language, or contain specialized terminology.

Question 329. Which Azure AI capability converts spoken audio into written text?

1. Speech-to-text
2. Text-to-speech
3. Key phrase extraction
4. Image analysis

Correct Answer: 1. Speech-to-text

Explanation:

Speech-to-text converts spoken language from an audio input into written text. This capability can support applications such as voice assistants, meeting transcription, call-center analysis, accessibility tools, and voice-controlled interfaces. Recognition quality can depend on factors such as language, dialect, microphone quality, background noise, speaking speed, and specialized vocabulary. Developers should therefore test speech recognition with representative recordings from the intended users and environment. Speech-to-text is different from text-to-speech, which performs the opposite transformation by generating spoken audio from written text. Azure AI Speech provides speech-related capabilities that can be integrated into many application scenarios.

Question 330. Which Azure AI capability converts written text into spoken audio?

1. Text-to-speech
2. Speech-to-text
3. OCR
4. Clustering

Correct Answer: 1. Text-to-speech

Explanation:

Text-to-speech converts written text into synthesized spoken audio. It can be used for accessibility features, virtual assistants, navigation systems, educational applications, and other solutions where users need information delivered through speech. Developers can select supported languages, locales, and voices based on the requirements of the application. Voice selection should consider pronunciation, clarity, audience, and the desired communication style. Specialized terminology should also be tested because pronunciation may require additional consideration. Text-to-speech is the reverse of speech-to-text, which converts spoken audio into written text. Both capabilities are available through Azure AI Speech for supported scenarios.

Question 331. Which Azure AI service can translate text between supported languages?

1. Azure AI Translator
2. Azure Machine Learning
3. Azure Blob Storage
4. Azure Monitor

Correct Answer: 1. Azure AI Translator

Explanation:

Azure AI Translator provides machine translation capabilities for converting text from one supported language to another. Translation can be integrated into applications that need to support multilingual users, translate documents or messages, or provide localized content. Developers should consider the languages and locales supported by the service and evaluate translation quality for their particular domain. Technical terminology, names, cultural context, and ambiguous phrases may require additional review. Machine translation can significantly reduce manual translation effort, but high-impact or highly specialized content may still require human review depending on the application’s requirements and consequences of incorrect translation.

Question 332. Which Azure AI service is designed to extract structured information from documents?

1. Azure AI Document Intelligence
2. Azure AI Translator
3. Azure DNS
4. Azure Firewall

Correct Answer: 1. Azure AI Document Intelligence

Explanation:

Azure AI Document Intelligence is designed to extract information from documents and convert unstructured or semi-structured content into structured data. It can support scenarios involving forms, invoices, receipts, identity-related documents, and other document types depending on the available models and application requirements. Extracted information can then be used by business applications or automated workflows. Developers should select prebuilt or custom models according to the document format and extraction requirements. Document quality, layout variation, handwriting, and unusual content can affect extraction accuracy, so solutions should be tested with representative documents before being used for important business processes.

Question 333. What does OCR primarily do when processing an image or scanned document?

1. Extracts printed or handwritten text from visual content
2. Predicts stock prices
3. Creates database indexes
4. Converts text into speech

Correct Answer: 1. Extracts printed or handwritten text from visual content

Explanation:

Optical Character Recognition, or OCR, identifies text contained within images or scanned documents and converts that visual text into machine-readable information. OCR is useful for digitizing documents, extracting information from images, enabling text search, and supporting downstream document-processing workflows. Recognition accuracy can depend on image quality, resolution, font style, layout, handwriting, lighting, and other factors. OCR is focused on recognizing text rather than fully understanding every aspect of a document’s meaning. For structured extraction from forms or invoices, additional document-analysis capabilities may be appropriate. Developers should evaluate OCR results using representative documents.

Question 334. Which Azure AI Vision capability identifies individual objects and their locations within an image?

1. Object detection
2. Text summarization
3. Language detection
4. Regression

Correct Answer: 1. Object detection

Explanation:

Object detection identifies objects within an image and provides information about where those objects are located. Unlike simple image classification, which assigns one or more labels to an image or category, object detection can identify multiple objects and their positions, often represented by bounding boxes. This makes it useful for scenarios such as inventory analysis, visual inspection, traffic monitoring, and image-based applications. Developers should evaluate detection performance using representative images and consider factors such as object size, lighting, image quality, and environmental conditions. The appropriate computer vision capability depends on whether the application needs labels, locations, segmentation, or other visual information.

Question 335. Which computer vision technique assigns a class label to an entire image?

1. Image classification
2. Speech recognition
3. Named entity recognition
4. Document translation

Correct Answer: 1. Image classification

Explanation:

Image classification assigns one or more categories to an image based on its visual content. For example, a classification model might determine whether an image belongs to categories such as dog, cat, vehicle, or landscape. It differs from object detection because classification does not primarily identify the precise location of each object in the image. The appropriate training data should represent the images and conditions expected in the intended application. Developers should evaluate the model using unseen images and consider accuracy across relevant categories. Poorly represented classes or differences between training and production images can affect classification performance.

Question 336. What is image segmentation designed to determine?

1. Which pixels belong to specific objects or regions
2. Which language a document uses
3. Which words are most important
4. Which speech sample is the loudest

Correct Answer: 1. Which pixels belong to specific objects or regions

Explanation:

Image segmentation identifies regions or pixels in an image that belong to particular objects or categories. This provides more detailed information than simply assigning a label to an entire image or drawing a bounding box around an object. Segmentation can be useful in scenarios where the exact shape or boundaries of an object matter, such as detailed visual inspection or specialized image analysis. The quality of segmentation depends on the training data, image conditions, and complexity of the objects being analyzed. Developers should determine whether their application actually requires pixel-level information before selecting segmentation over simpler vision techniques.

Question 337. Which Azure AI Search feature can improve retrieval when users search for concepts rather than exact words?

1. Vector search
2. Text-to-speech
3. Speech recognition
4. Model deployment

Correct Answer: 1. Vector search

Explanation:

Vector search uses numerical representations of content to identify items that are semantically similar to a query. This is useful when users express a concept using different words from those contained in the stored documents. For example, a user might ask about changing a password while a document describes updating account credentials. Keyword-only search may have difficulty recognizing that relationship, while vector-based retrieval can identify semantic similarity. Azure AI Search supports vector search scenarios that can be combined with keyword retrieval. Developers should use appropriate embeddings and evaluate retrieval quality using representative queries and relevant source documents.

Question 338. What is retrieval-augmented generation primarily used for?

1. Combining retrieved information with generative AI responses
2. Replacing all databases with neural networks
3. Converting images into speech
4. Removing all source documents

Correct Answer: 1. Combining retrieved information with generative AI responses

Explanation:

Retrieval-augmented generation, or RAG, combines information retrieval with a generative AI model. When a user submits a question, the application first retrieves relevant information from a knowledge source such as documents or a search index. The retrieved content is then provided as context to the generative model, which uses that information when constructing a response. This approach can be useful when applications need answers based on organizational or domain-specific information. RAG does not automatically guarantee factual accuracy, so developers should evaluate retrieval quality, source relevance, permissions, prompt handling, and generated responses.

Question 339. Which practice can help prevent an AI application from exposing information a user is not authorized to access?

1. Applying access controls to retrieved data
2. Giving every user administrator permissions
3. Disabling authentication
4. Including every document in every prompt

Correct Answer: 1. Applying access controls to retrieved data

Explanation:

Access controls help ensure that users can retrieve only information they are authorized to access. This is particularly important in AI applications using enterprise search or retrieval-augmented generation because a model may otherwise receive sensitive documents as context and potentially include information from them in a response. Permissions should therefore be enforced at the application and data-retrieval layers rather than relying solely on the language model to decide what information is allowed. Identity management, authorization checks, secure data sources, and careful logging can help protect sensitive information. AI applications should treat access control as a core security requirement.

Question 340. Why should AI systems be tested with edge cases before production deployment?

1. To identify unexpected behavior under unusual or difficult conditions
2. To guarantee zero maintenance
3. To eliminate all model updates
4. To increase database storage automatically

Correct Answer: 1. To identify unexpected behavior under unusual or difficult conditions

Explanation:

Edge-case testing evaluates how an AI system behaves when it receives unusual, ambiguous, incomplete, unexpected, or difficult inputs. Models may perform well on common examples but behave differently when conditions fall outside the most frequently observed patterns. Testing edge cases can reveal weaknesses in model accuracy, safety, robustness, input handling, and application logic before users encounter them in production. Examples should be relevant to the intended scenario and can include unusual language, unexpected document formats, rare classes, incomplete information, or adversarial inputs. Identifying these limitations early allows developers to add safeguards, improve data, adjust prompts, or refine the system.