View Full Microsoft AI-102 Exam Dumps and Practice Test Dumps.

 

Question 81

Which capability allows an AI application to determine the category of an incoming support request?

1. OCR
2. Text classification
3. Face detection
4. Speech synthesis

Correct Answer: 2

Explanation

Text classification assigns text to predefined categories based on its content. In a support application, incoming requests might be classified as billing, technical support, account access, or product inquiries. The classification result can then be used to route the request to the appropriate team or workflow. OCR extracts text from images, face detection analyzes visual content, and speech synthesis converts text into audio. A classification solution should be trained or configured with representative examples and evaluated against realistic data. This helps ensure that categories are meaningful and that the application can handle variations in how users describe their requests.

Question 82

Which Azure service is designed to provide conversational language understanding for applications?

1. Azure AI Search
2. Azure AI Vision
3. Azure AI Language
4. Azure Storage

Correct Answer: 3

Explanation

Azure AI Language provides natural language processing capabilities that can help applications understand and analyze human language. Depending on the workload, capabilities include sentiment analysis, key phrase extraction, named entity recognition, language detection, summarization, and text classification. Azure AI Search focuses primarily on search and retrieval, Azure AI Vision analyzes visual content, and Azure Storage provides data storage rather than language understanding. When an application needs to interpret textual user input or analyze natural-language content, Azure AI Language is a relevant service. Developers should select the specific language capability based on the application’s functional requirements.

Question 83

A developer needs to identify the location of several cars within a photograph. Which capability is appropriate?

1. Image classification
2. Text summarization
3. Object detection
4. Language detection

Correct Answer: 3

Explanation

Object detection identifies objects within an image and provides information about where those objects appear. For a photograph containing several cars, an object detection model can identify the cars and return their corresponding locations, typically represented by bounding regions. Image classification instead assigns labels to an image without necessarily locating individual objects. Text summarization and language detection operate on textual information. Object detection is therefore appropriate when the application needs both identification and localization of visual objects. Developers should also evaluate detection performance using representative images, especially when lighting, camera angles, or object sizes vary.

Question 84

What is the purpose of a vector embedding in semantic search?

1. Store passwords
2. Represent content numerically
3. Compress audio
4. Configure DNS

Correct Answer: 2

Explanation

A vector embedding represents content as a numerical vector that captures meaningful characteristics of the original information. In semantic search, text or other supported content can be converted into embeddings and stored in a vector index. A query can then be embedded and compared with stored vectors to identify semantically similar content. This allows retrieval even when the query and document use different words to express related concepts. Embeddings are not intended for password storage, audio compression, or DNS configuration. Their primary role is to enable mathematical comparison of content based on learned semantic representations.

Question 85

Which feature can be used to translate text from one supported language into another?

1. Object detection
2. Translator
3. OCR
4. Face analysis

Correct Answer: 2

Explanation

Azure AI Translator provides machine translation for supported languages. Applications can use it to translate written content in scenarios such as multilingual customer support, websites, document workflows, and communication systems. The source text is submitted to the translation service, which returns content in the requested target language. Object detection and face analysis work with images, while OCR extracts text from visual content rather than translating it. Translator can also be combined with other services; for example, OCR can extract text from an image and Translator can then translate the extracted text into another language.

Question 86

A generative AI application needs current product information from a company database. What should the application do?

1. Use only the model’s training knowledge
2. Ignore the database
3. Retrieve current information
4. Remove grounding

Correct Answer: 3

Explanation

When information changes frequently, the application should retrieve current information from an authoritative source rather than relying exclusively on the model’s general knowledge. The retrieved data can then be supplied to the generative model as grounding context. For example, current product prices, availability, or specifications can be queried from a database before generating a response. This approach helps reduce the risk of outdated answers. The model’s training knowledge may still provide general context, but it should not be treated as a reliable source for rapidly changing organizational information. Retrieval therefore plays an important role in current-data scenarios.

Question 87

Which component of Azure AI Search contains searchable representations of indexed content?

1. Skillset
2. Data source
3. Index
4. Indexer

Correct Answer: 3

Explanation

An Azure AI Search index contains the searchable representation of content that has been ingested into the search service. It defines fields and supports capabilities such as keyword, filtering, sorting, and, when configured, vector and semantic search. A data source identifies where content originates, while an indexer can automate the process of retrieving content from supported sources and placing it into the index. A skillset defines enrichment operations that can transform or augment content during indexing. Understanding these components is essential when designing search and knowledge-mining pipelines for AI applications.

Question 88

Which approach is most suitable when an AI assistant must use internal documents while answering user questions?

1. Image classification
2. RAG
3. Speech synthesis
4. Face detection

Correct Answer: 2

Explanation

Retrieval-augmented generation, or RAG, allows an AI assistant to retrieve relevant information from an external knowledge source and provide that information to a generative model as context. For internal documents, the organization can index approved content and retrieve relevant passages based on each user’s question. The model can then generate a response grounded in the retrieved material. Image classification, speech synthesis, and face detection address unrelated workloads. A RAG implementation should also include access controls, retrieval evaluation, appropriate prompting, and monitoring so that users receive information from authorized and relevant sources.

Question 89

Which capability is most appropriate for extracting names of people and organizations from text?

1. Speech recognition
2. Image segmentation
3. Named entity recognition
4. Text-to-speech

Correct Answer: 3

Explanation

Named entity recognition identifies entities such as people, organizations, locations, dates, and other supported categories within text. This capability can transform unstructured language into useful structured information for search, analytics, routing, and knowledge-mining scenarios. For example, a contract-processing application can identify company names and individuals mentioned in the document. Speech recognition converts audio into text, image segmentation analyzes visual regions, and text-to-speech creates spoken output. Named entity recognition is therefore the appropriate language-processing capability when an application needs to identify specific entity types within written content.

Question 90

A developer wants to ensure an AI response follows a specific role and set of behavioral instructions. What should be configured?

1. Storage policy
2. System instructions
3. Network route
4. Image format

Correct Answer: 2

Explanation

System instructions can define the role, behavior, constraints, and communication requirements that should guide a generative AI model. For example, an application might instruct the model to act as a customer-support assistant, use concise language, rely on supplied context, and avoid answering questions outside a defined scope. These instructions are part of prompt design and can strongly influence model behavior. Storage policies, network routes, and image formats do not directly establish conversational behavior. Developers should test system instructions with varied inputs because instructions alone cannot guarantee perfect adherence in every situation.

Question 91

Which Azure AI Vision capability can identify multiple objects and their locations in an image?

1. Object detection
2. Language detection
3. Key phrase extraction
4. Speech translation

Correct Answer: 1

Explanation

Object detection identifies objects within an image and provides information about their locations. This allows an application to determine not only that certain objects exist but also where they appear within the visual content. For example, a warehouse application could detect boxes, equipment, or other objects in photographs. Language detection and key phrase extraction operate on text, while speech translation processes spoken language. Object detection is therefore the appropriate computer vision capability when the solution needs to locate multiple visual objects. Accuracy should be tested against realistic images representing the conditions expected in production.

Question 92

Why might an AI application use semantic search instead of only keyword search?

1. It can understand conceptual relationships
2. It removes all security requirements
3. It guarantees perfect results
4. It eliminates indexing

Correct Answer: 1

Explanation

Semantic search can identify content based on meaning and conceptual relationships rather than depending entirely on exact keyword matches. This can improve retrieval when users and documents use different words to describe related concepts. For example, a query about automobile servicing may retrieve documents discussing vehicle maintenance even if the exact phrase in the query is not present. Semantic search does not guarantee perfect results or eliminate the need for indexing and security controls. It should be evaluated using representative queries and documents so that developers can determine whether the resulting relevance meets the application’s requirements.

Question 93

A company wants to extract information from receipts without developing a model from scratch. What should it consider?

1. Speech translation
2. A suitable prebuilt Document Intelligence model
3. Semantic ranking
4. Text-to-speech

Correct Answer: 2

Explanation

A suitable prebuilt Document Intelligence model can reduce development effort when the organization’s documents match a supported predefined scenario. Receipt processing is a common document-analysis workload for which structured information such as merchant names, transaction dates, totals, and other fields may be extracted. A prebuilt model can provide a starting point without requiring the organization to develop a custom model from scratch. Speech translation, semantic ranking, and text-to-speech address different workloads. The extracted results should still be tested against representative receipts because document layouts, image quality, and field visibility can affect extraction accuracy.

Question 94

Which capability is useful for generating a concise version of a long document?

1. Object detection
2. Face detection
3. Summarization
4. Speech recognition

Correct Answer: 3

Explanation

Summarization generates a shorter representation of a longer piece of content while attempting to retain its important information. It can be useful for reports, customer feedback, meeting notes, articles, and other documents where users need to understand the main points quickly. Object detection and face detection analyze visual content, while speech recognition converts spoken audio into text. A summarization solution should be evaluated to ensure that important facts are retained and that the resulting summary is appropriate for the intended use. For sensitive or high-impact content, human review may also be appropriate.

Question 95

What should be used to determine whether generated content violates configured safety requirements?

1. Content safety controls
2. DNS settings
3. Database indexing
4. Image resolution

Correct Answer: 1

Explanation

Content safety controls can help detect or manage potentially harmful content generated by an AI application. Depending on the service and configuration, safety mechanisms can evaluate prompts and model responses against supported categories and thresholds. These controls provide an important protection layer for generative AI applications, particularly those exposed to external users. DNS settings, database indexing, and image resolution do not determine whether generated content meets safety requirements. Developers should combine content safety with appropriate prompts, access controls, testing, monitoring, and application-specific safeguards rather than relying on a single control.

Question 96

Which Azure AI Speech capability converts written text into spoken audio?

1. Speech recognition
2. Language detection
3. Text-to-speech
4. OCR

Correct Answer: 3

Explanation

Text-to-speech converts written text into synthesized spoken audio. Azure Speech can provide this capability for applications such as virtual assistants, accessibility tools, interactive systems, and automated announcements. The application supplies text and receives audio output using the selected supported voice and configuration. Speech recognition performs the opposite general direction by converting spoken audio into text. Language detection identifies the language of text, while OCR extracts text from images. Text-to-speech is therefore the appropriate capability when an application has written content and needs to present that information through natural-sounding spoken output.

Question 97

Which search strategy combines lexical matching with semantic vector retrieval?

1. Exact search
2. Hybrid search
3. Image search
4. Audio search

Correct Answer: 2

Explanation

Hybrid search combines traditional lexical or keyword-based retrieval with vector-based semantic retrieval. Keyword matching can provide strong results when specific terms are important, while vector retrieval can identify conceptually related content even when the wording differs. Combining both approaches can improve retrieval quality for enterprise search and retrieval-augmented generation scenarios. Exact search alone does not provide the same semantic capability, while image and audio search describe different content modalities. Engineers should evaluate hybrid search using representative queries and documents and tune the retrieval configuration according to the needs of the application.

Question 98

What is a key purpose of evaluating an AI application with representative test data?

1. To increase storage
2. To disable monitoring
3. To identify performance and quality issues
4. To remove authentication

Correct Answer: 3

Explanation

Representative test data allows developers to evaluate whether an AI application performs appropriately under conditions similar to its expected production workload. Testing can reveal issues with accuracy, relevance, grounding, safety, latency, or other application-specific requirements. Using only a few ideal examples may hide problems that occur with unusual wording, incomplete information, different document formats, or edge cases. Increasing storage, disabling monitoring, and removing authentication do not improve evaluation quality. A representative evaluation set provides evidence that can guide prompt refinement, model selection, retrieval configuration, and other improvements before and after deployment.

Question 99

An AI agent needs to call an external service to retrieve customer order information. What should be configured?

1. A tool or connected capability
2. Image captioning
3. OCR only
4. Sentiment analysis

Correct Answer: 1

Explanation

An AI agent can use a tool or connected capability to interact with an external service and retrieve information such as customer order details. The tool provides a defined interface that allows the application to perform an approved operation rather than expecting the language model to access external systems directly. Tool access should be secured with appropriate authentication and authorization, and inputs should be validated before an operation is performed. Image captioning, OCR, and sentiment analysis solve different problems. Tool integration is therefore a key mechanism for extending an agent beyond generating responses from its own conversational context.

Question 100

Which practice helps protect sensitive information used by an AI application?

1. Put credentials in prompts
2. Share secrets with users
3. Apply least-privilege access
4. Disable authentication

Correct Answer: 3

Explanation

Least-privilege access limits identities and application components to only the permissions required to perform their intended tasks. This reduces the potential impact if credentials, applications, or connected resources are compromised. Sensitive credentials should not be placed in prompts or exposed to users, and authentication should remain enabled for protected resources. In addition to access control, secure secret management, encryption, logging, monitoring, and appropriate data-handling practices should be considered. Least privilege is therefore an important security principle for AI applications that interact with private data, models, databases, storage accounts, or external services.