AI Associate [AI-102] 3 — Questions and Answers
Question 1: What is 'grounding' in the context of Retrieval-Augmented Generation (RAG) with Azure OpenAI?
- Reducing model hallucinations by providing retrieved context as part of the prompt (Correct answer)
- Fine-tuning a model on domain-specific data
- Setting the system prompt to restrict model behavior
- Applying content filters to model outputs
Correct answer: Reducing model hallucinations by providing retrieved context as part of the prompt
Grounding anchors model responses to retrieved external documents, reducing hallucinations by giving the model factual context in the prompt.
Question 2: In Azure Cognitive Search, which analyzer type should you choose to support language-specific stemming for French documents?
- Standard Lucene analyzer
- Keyword analyzer
- Language analyzer (fr.microsoft) (Correct answer)
- Pattern analyzer
Correct answer: Language analyzer (fr.microsoft)
Language-specific analyzers like fr.microsoft apply stemming, stop-word removal, and tokenization rules tailored to a specific language.
Question 3: Which AI-102 concept describes splitting a large document into smaller pieces before embedding them for vector search?
- Tokenization
- Chunking (Correct answer)
- Normalization
- Vectorization
Correct answer: Chunking
Chunking divides large documents into manageable segments so each chunk can be embedded and retrieved independently with higher relevance.
Question 4: You need to identify named entities like people, organizations, and locations from customer feedback. Which Azure AI Language feature should you use?
- Key phrase extraction
- Named Entity Recognition (NER) (Correct answer)
- Sentiment analysis
- Language detection
Correct answer: Named Entity Recognition (NER)
Named Entity Recognition identifies and categorizes entities such as persons, organizations, and locations within unstructured text.
Question 5: What is the role of an 'enrichment pipeline' in Azure Cognitive Search?
- It synchronizes the search index with a SQL database
- It applies AI skills during indexing to extract and transform content (Correct answer)
- It caches query results for faster responses
- It encrypts index data before storage
Correct answer: It applies AI skills during indexing to extract and transform content
An enrichment pipeline (skillset) applies AI skills—OCR, entity extraction, translation—to raw content during indexing to augment the searchable index.
Question 6: Which deployment option allows you to host an Azure AI model locally on a device without requiring internet connectivity?
- Azure Kubernetes Service deployment
- Azure IoT Edge with AI modules (Correct answer)
- Azure Functions consumption plan
- Azure Container Apps
Correct answer: Azure IoT Edge with AI modules
Azure IoT Edge allows deploying containerized AI models as edge modules that run locally on devices without continuous cloud connectivity.
Question 7: In Azure Machine Learning, what is the purpose of 'model registration'?
- Deploying the model to an online endpoint
- Versioning and storing trained models in a central repository for tracking and reuse (Correct answer)
- Scheduling automated retraining jobs
- Configuring compute clusters for training
Correct answer: Versioning and storing trained models in a central repository for tracking and reuse
Model registration stores versioned model artifacts in the Azure ML model registry, enabling tracking, auditing, and controlled deployment.
AI Associate [AI-102] 3
What is 'grounding' in the context of Retrieval-Augmented Generation (RAG) with Azure OpenAI?