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

 

Question 141

Which Azure AI Vision feature can identify objects and their locations within an image?

1. Image captioning
2. Object detection
3. Face verification
4. Image translation

Correct Answer: 2

Explanation

Object detection identifies individual objects within an image and provides information about where those objects appear. Unlike image classification, which assigns labels to an entire image, object detection can identify multiple objects and associate each detected object with a location, commonly represented by a bounding box. This makes the capability useful for scenarios such as inventory monitoring, manufacturing inspection, retail analysis, and automated visual processing. Developers should select object detection when an application needs both the type of object and its position rather than simply determining the overall category of an image.

Question 142

An application needs to convert spoken English into written text in real time. Which Azure AI service should be used?

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

Correct Answer: 1

Explanation

Azure AI Speech provides speech-to-text capabilities that can convert spoken language into written text. Real-time speech recognition is useful for applications such as voice assistants, meeting transcription, accessibility tools, call-center solutions, and interactive applications. The service can process audio input and return recognized text as speech is detected. Developers can configure language and recognition settings according to the application requirements. When building a complete voice application, speech-to-text can also be combined with Azure AI Language or generative AI services to interpret the recognized text and produce an appropriate response.

Question 143

A document-processing application must determine whether a submitted document belongs to an invoice, receipt, or contract category before extracting fields. Which approach is appropriate?

1. Document classification
2. Speech translation
3. Image generation
4. Sentiment analysis

Correct Answer: 1

Explanation

Document classification can determine the category or type of a document before additional processing occurs. In an enterprise workflow, a system might first classify incoming files as invoices, receipts, contracts, applications, or other supported categories. The classification result can then determine which extraction model or processing workflow should be applied. This is especially useful when documents arrive through a common intake channel and their types are unknown in advance. Separating classification from extraction can make the overall pipeline easier to manage and can help route documents to the appropriate downstream processing service.

Question 144

Which Azure AI Search feature allows an application to present counts of documents grouped by a field such as department or document type?

1. Vectorization
2. Faceting
3. OCR
4. Semantic captions

Correct Answer: 2

Explanation

Faceting allows Azure AI Search applications to group search results according to selected fields and provide counts for each category. For example, a knowledge portal could show the number of matching documents belonging to Finance, Human Resources, Sales, or Engineering. Users can then select a facet to narrow the search results. Faceting is particularly useful in enterprise search interfaces because it gives users an organized way to explore large result sets. Fields used for facets must be configured appropriately in the search index. Faceting complements retrieval and filtering rather than replacing them.

Question 145

A developer wants to make a generative AI application answer questions using a company’s private documents instead of relying only on model training data. Which architecture is appropriate?

1. RAG
2. OCR only
3. Image classification
4. Speech synthesis

Correct Answer: 1

Explanation

Retrieval-augmented generation, or RAG, allows a generative AI application to retrieve relevant information from an external knowledge source and provide that information to the model as context. A company’s private documents can be processed, indexed, and represented for retrieval. When a user asks a question, the system retrieves relevant content and includes it in the model request. This helps the application answer using current or organization-specific information without requiring the base model to be retrained for every document update. Proper access control, retrieval quality, and grounding instructions remain important for reliable enterprise applications.

Question 146

Which Azure AI Language capability identifies the emotional polarity of text as positive, negative, or neutral?

1. Key phrase extraction
2. Sentiment analysis
3. Language detection
4. Entity linking

Correct Answer: 2

Explanation

Sentiment analysis determines the emotional polarity expressed in text, commonly identifying positive, negative, or neutral sentiment. Depending on the service capabilities and configuration, sentiment analysis can also provide more detailed information about individual sentences or aspects of the text. Businesses can use this capability to analyze customer feedback, reviews, survey responses, support interactions, and other textual information. Sentiment analysis does not simply search for positive or negative words; the language model evaluates the text in context. Applications should validate results against representative business data before relying on them for important decisions.

Question 147

A developer wants a generative AI application to follow a fixed set of behavioral rules throughout conversations. Where should these high-level instructions primarily be defined?

1. System instructions
2. User profile image
3. Search index
4. Audio transcript

Correct Answer: 1

Explanation

System instructions define high-level behavioral guidance for a generative AI model or agent. They can specify requirements such as the assistant’s role, response style, task boundaries, or rules for using available information. For example, an enterprise assistant might be instructed to answer using retrieved company documentation and state when information is unavailable. System instructions should be written clearly and should not be treated as a replacement for application-level security controls. Authorization, data access, validation, and tool permissions must still be enforced outside the model because natural-language instructions alone cannot guarantee security.

Question 148

An application must identify text such as people’s names, organizations, locations, and dates. Which Azure AI Language capability is designed for this?

1. Named entity recognition
2. Speech synthesis
3. Image analysis
4. Key phrase ranking

Correct Answer: 1

Explanation

Named entity recognition identifies entities within unstructured text and categorizes them into supported entity types. Examples can include people, organizations, locations, dates, quantities, and other recognized categories. This capability can help applications transform unstructured documents into more structured information. For example, a customer-support system could extract organization and location names from incoming requests for routing or analysis. Standard named entity recognition is useful when the required entity types are already supported. If an organization needs highly specialized entity categories, custom named entity recognition can be considered instead.

Question 149

Which Azure AI Search configuration determines which fields are used by semantic ranking for a semantic query?

1. Semantic configuration
2. Data source connection
3. Storage account
4. Speech profile

Correct Answer: 1

Explanation

A semantic configuration in Azure AI Search identifies the fields that should be considered by semantic ranking. It can specify important content such as title fields, prioritized content fields, and other relevant searchable information. This gives the semantic ranker a structured understanding of which portions of indexed documents are most useful for interpreting a query. Proper configuration can improve the quality of semantic results and captions. The semantic configuration works with the search index and query process; it does not replace the index itself or perform document ingestion.

Question 150

A company wants to use a model from the Microsoft Foundry model catalog in an application. What must generally happen before the application can send inference requests?

1. Deploy the model
2. Delete the project
3. Disable authentication
4. Remove the model

Correct Answer: 1

Explanation

A model selected from the Microsoft Foundry model catalog generally needs to be deployed or otherwise made available through an appropriate deployment mechanism before an application can send inference requests to it. The deployment provides an endpoint or configured access path through which the application can interact with the model. Developers should also configure authentication, authorization, and appropriate resource settings. Model selection should consider capabilities, supported modalities, performance, cost, and application requirements. The deployment process therefore forms an important bridge between choosing a model during development and consuming it from an application.

Question 151

Which Azure AI Document Intelligence capability is most appropriate when an organization needs to extract specialized fields from its own document format?

1. Custom extraction model
2. Speech recognizer
3. Sentiment model
4. Image captioner

Correct Answer: 1

Explanation

A custom extraction model in Azure AI Document Intelligence is appropriate when an organization needs to extract fields from documents with specialized layouts or business-specific information. Developers can provide representative labeled documents and define the fields that should be extracted. The trained model can then process similar documents and return structured field information. This differs from prebuilt models, which are designed for commonly encountered document types and standardized fields. Custom extraction is useful when an organization’s documents contain unique fields or structures that are not adequately covered by the available prebuilt Document Intelligence models.

Question 152

A voice assistant needs to respond with synthesized speech that sounds natural rather than displaying text. Which Azure AI capability should be used?

1. Text-to-speech
2. Vector search
3. OCR
4. Document classification

Correct Answer: 1

Explanation

Text-to-speech converts written text into spoken audio using Azure AI Speech. It can be used to provide audible responses in voice assistants, accessibility applications, automated customer-service systems, navigation tools, and other conversational experiences. The application can generate or retrieve a text response and then submit that text to the speech synthesis service. Voice selection and supported synthesis controls can be configured according to the desired experience. When combined with speech-to-text, text-to-speech can form one half of a complete voice interaction pipeline in which users speak to an application and receive spoken responses.

Question 153

A search application should return documents that contain the exact terms from a user’s query even when their semantic meaning is not similar. Which retrieval method is appropriate?

1. Keyword search
2. Vector-only search
3. Image analysis
4. Speech recognition

Correct Answer: 1

Explanation

Keyword search is appropriate when exact terms or lexical matches are important to the retrieval task. Azure AI Search supports traditional text retrieval that can identify documents containing relevant query terms even when their broader semantic meaning is not close to another phrase. This can be particularly useful for product codes, names, identifiers, technical terms, and exact terminology. Vector search approaches queries according to embedding similarity, while keyword search focuses on textual matching. Many enterprise applications benefit from combining both approaches through hybrid search so that exact and semantic retrieval signals can complement each other.

Question 154

An AI solution must identify personally identifiable information in customer messages before storing the processed text. Which Azure AI Language capability can help?

1. PII detection
2. Image tagging
3. Speech translation
4. Object tracking

Correct Answer: 1

Explanation

PII detection in Azure AI Language can identify supported types of personally identifiable information in text. This can help organizations locate sensitive values such as names, addresses, phone numbers, or other recognized personal information before text is stored or processed further. Depending on the application, identified information can then be redacted, masked, or handled according to organizational privacy requirements. PII detection is a processing capability rather than a complete privacy strategy. Applications should also implement appropriate access controls, retention policies, encryption, auditing, and other safeguards for sensitive information.

Question 155

Which Azure AI Search field type is designed to store numerical embedding values for vector retrieval?

1. Vector field
2. Image field
3. Boolean-only field
4. Date-only field

Correct Answer: 1

Explanation

A vector field stores numerical vector representations that can be used for vector search in Azure AI Search. These vectors are commonly generated by an embedding model from text, images, or other supported content. The field is configured with the appropriate dimensions and vector search settings so that queries can compare the query vector with stored vectors. Vector fields are usually accompanied by metadata and source-content fields that provide information needed after retrieval. Properly matching the vector dimensions and embedding model configuration is important because incompatible vector representations can prevent effective similarity searching.

Question 156

A developer needs to extract the most important concepts from a collection of customer comments without assigning predefined categories. Which capability should be considered?

1. Key phrase extraction
2. Custom text classification
3. Face detection
4. Speech synthesis

Correct Answer: 1

Explanation

Key phrase extraction identifies important terms and concepts within unstructured text. Unlike custom text classification, it does not require predefined categories such as Billing or Technical Support. This makes it useful for quickly discovering themes in customer comments, survey responses, reviews, or other large collections of text. The extracted phrases can support indexing, analytics, summarization workflows, or subsequent processing. Developers should remember that key phrases represent important concepts rather than providing a complete summary of the source text. Results should therefore be interpreted according to the specific requirements of the application.

Question 157

A developer is testing a generative AI application and wants to trace the sequence of model and tool calls made during a workflow. What capability is useful?

1. Tracing
2. OCR
3. Faceting
4. Image classification

Correct Answer: 1

Explanation

Tracing helps developers inspect the sequence and details of operations performed during an AI workflow. In a generative AI application, a trace can help reveal how prompts, model calls, retrieval operations, and tools interact during execution. This information can be useful when diagnosing incorrect results, unexpected tool usage, latency problems, or failures in multi-step workflows. Tracing should be used alongside evaluation and monitoring rather than as a replacement for them. When traces contain sensitive prompts or retrieved content, organizations should also apply appropriate privacy, access-control, and data-retention policies.

Question 158

An application must translate spoken French into written English. Which processing sequence is most appropriate?

1. OCR then search
2. Speech translation
3. Object detection then OCR
4. Sentiment then classification

Correct Answer: 2

Explanation

Speech translation is designed for scenarios where spoken language needs to be translated into another language. Azure AI Speech can process spoken input, recognize the source language, and provide translated output in the target language. This is different from ordinary speech-to-text, which converts speech into text without necessarily translating it. Speech translation can support multilingual meetings, customer-service interactions, travel applications, and accessibility scenarios. Developers should configure the appropriate source and target languages and test the system using representative audio because pronunciation, background noise, accents, and speaking conditions can affect recognition quality.

Question 159

A developer wants an AI Search query to return only documents where the accessLevel field equals “internal”. Which search capability should be used?

1. Filtering
2. Semantic captions
3. Image analysis
4. Text-to-speech

Correct Answer: 1

Explanation

Filtering allows Azure AI Search queries to restrict results according to structured field values. If an index contains an accessLevel field configured as filterable, a query can restrict results to documents whose value matches a specified condition, such as internal. Filters are particularly important in enterprise search because they can narrow retrieval according to metadata such as department, document type, region, status, or access category. Developers should design filterable fields carefully and ensure that application-level authorization is also enforced. A search filter should not be considered a substitute for comprehensive access-control mechanisms.

Question 160

A team wants to verify that an AI application remains accurate after a new model version is deployed. What should the team perform?

1. Regression evaluation
2. Image resizing
3. OCR conversion
4. Voice selection

Correct Answer: 1

Explanation

Regression evaluation compares the behavior of a new application or model version against established expectations or a previous version. After deploying a new model, developers can run representative test cases to determine whether important capabilities remain correct and whether previously successful scenarios have degraded. This is especially important for generative AI systems because changes in models, prompts, retrieval components, or tools can alter outputs in unexpected ways. A regression evaluation dataset should represent important production scenarios and include measurable criteria where possible. Results can help teams identify issues before or after broader deployment.