AI-900 Exam Prep
AI-900 Exam Glossary - 39 Terms
Search the terminology pack for Microsoft Azure AI Fundamentals. Use these definitions with the study guide and practice questions.
A
- Accountability
- A responsible AI principle requiring that AI systems be subject to human oversight, audit, and review.
- AI Impact Assessment
- A structured evaluation used to document an AI system’s purpose, expected use, risks, and potential harms.
- Auditability
- The ability to examine and review an AI system’s behavior, decisions, and processes.
- Azure AI Question Answering
- An Azure AI service used to create knowledge bases from FAQs, documents, and other content to answer user questions.
- Azure AI Search
- An Azure service for indexing, searching, and retrieving information from structured and unstructured content.
- Azure Bot Service
- An Azure service for building, deploying, and managing chatbots that interact with users through conversational interfaces.
- Azure OpenAI Service
- Microsoft’s managed Azure service for deploying, hosting, and using OpenAI generative AI models.
C
- Chatbot
- A software application that simulates conversation with users through text or voice interactions.
- Client library
- A language-specific software package that simplifies calling and integrating with a service API.
- Computer vision
- An AI field and Azure capability focused on extracting meaning and insights from images and visual data.
- Content filters
- Safety mechanisms that detect, block, or flag harmful or unsafe prompts and model outputs.
- Conversational AI
- AI systems designed to support natural dialogue between humans and machines using language understanding and response generation.
D
- Data source
- An external repository or storage location from which data is retrieved for processing or indexing.
E
- Entity
- A specific piece of information extracted from text, such as a date, location, or product name.
F
- FAQ document
- A document containing frequently asked questions and answers that can be imported into a knowledge base.
- Feature
- An input variable or attribute used by a machine learning model to make predictions.
G
- Generative AI
- AI systems that create new content such as text, images, or code from prompts or input data.
- GPT
- A family of generative pretrained transformer models used for language tasks such as summarization and text generation.
- Grayscale image
- An image represented by a single channel of pixel intensity values rather than multiple color channels.
I
- Indexer
- A component in Azure AI Search that automatically pulls data from supported data sources and populates a search index.
- Intent
- The goal or action a user wants to perform, as inferred from an utterance in a language understanding system.
K
- Knowledge base
- A collection of questions, answers, and supporting content used by AI systems to provide relevant responses.
L
- Language detection
- The process of identifying the language of a given text input.
- LUIS
- Language Understanding Intelligent Service, a tool used to build applications that identify user intents and extract entities from utterances.
M
- Machine learning model
- A trained mathematical model that learns patterns from data to make predictions or decisions.
N
- NaN
- Not a Number; a value that may be returned when a valid numeric score cannot be determined.
- Natural language understanding
- The capability of an AI system to interpret and extract meaning from human language input.
- Neural network
- A computational model that learns mappings from inputs to outputs by adjusting weighted connections during training.
P
- Privacy and security
- A responsible AI principle focused on protecting data from unauthorized access, misuse, or loss.
- Prompt
- The input or instruction provided to a generative AI model to guide its output.
- Pull model
- A data ingestion approach in which a service retrieves data from an external source rather than receiving pushed data.
R
- Regression
- A machine learning technique used to predict continuous numerical values such as prices or future measurements.
- Response generation
- The process of creating a relevant natural-language reply based on user input or context.
- Responsible AI
- A framework for designing and deploying AI systems in ways that are fair, safe, transparent, and accountable.
- REST API
- An application programming interface that enables access to services over HTTP using standard web methods.
S
- Search index
- A structured collection of searchable content used by a search service to enable fast information retrieval.
- Summarization
- The task of generating a concise version of a longer piece of text while preserving key meaning.
T
- Target variable
- The output value a machine learning model is trained to predict.
U
- Utterance
- A phrase or sentence submitted by a user to a conversational AI or language understanding system.
About These Definitions
These definitions are loaded from the shared release pack. Use them with the study guide and practice questions to connect vocabulary to exam scenarios.