Exam Domain
A major topic area covered by the exam.
Getting Started: Navigating the AI-900 Exam
Free knowledge base
Microsoft Azure AI Fundamentals (AI-900) — key terms, tricks & tips
Everything from the course in one searchable place: 232 entries. Use it to review before a practice test or look up a word you forgot.
232 results
Weighting
The percentage of questions from a specific domain.
Getting Started: Navigating the AI-900 Exam
Scaled Score
A converted raw score to ensure fairness across exams.
Getting Started: Navigating the AI-900 Exam
Skills Outline
Official document detailing all exam objectives.
Getting Started: Navigating the AI-900 Exam
Microsoft Learn
Free, self-paced online learning platform from Microsoft.
Getting Started: Navigating the AI-900 Exam
Multiple-choice
Question format with options, one or more correct.
Getting Started: Navigating the AI-900 Exam
Passing Score
Minimum score required to pass the certification exam.
Getting Started: Navigating the AI-900 Exam
Understanding the AI-900 Exam Structure
To remember the main AI-900 domains: 'CML-CNC'. Core AI, Machine Learning, Computer Vision, Natural Language, Conversational AI.
Getting Started: Navigating the AI-900 Exam
Understanding the AI-900 Exam Structure
The AI-900 exam requires a scaled score of 700 or greater to pass. Memorize the five main domain categories and their approximate percentage weightings to prioritize your study.
Getting Started: Navigating the AI-900 Exam
Understanding the AI-900 Exam Structure
Ignoring the official exam skills outline, leading to studying irrelevant topics or missing key ones.
Getting Started: Navigating the AI-900 Exam
Understanding the AI-900 Exam Structure
Not understanding the scaled scoring system and incorrectly assuming 70% correct answers guarantees a pass.
Getting Started: Navigating the AI-900 Exam
Understanding the AI-900 Exam Structure
Failing to practice different question formats, which can lead to confusion or slower response times during the actual exam.
Getting Started: Navigating the AI-900 Exam
Exam Skills Outline
Document detailing topics covered and their weighting on an exam.
Getting Started: Navigating the AI-900 Exam
Active Learning
Engaging with material through summarizing, teaching, or practice.
Getting Started: Navigating the AI-900 Exam
Study Schedule
A structured plan allocating time for specific study topics.
Getting Started: Navigating the AI-900 Exam
Practice Questions
Sample questions used to test knowledge and prepare for exam format.
Getting Started: Navigating the AI-900 Exam
Azure Documentation
Official technical guides and reference material for Azure services.
Getting Started: Navigating the AI-900 Exam
Knowledge Checks
Short quizzes within learning modules to assess understanding.
Getting Started: Navigating the AI-900 Exam
Effective Study Strategies for AI-900
LEARN: L-Leverage official resources, E-Establish a schedule, A-Active learning, R-Review regularly, N-Nail exam day prep!
Getting Started: Navigating the AI-900 Exam
Effective Study Strategies for AI-900
The AI-900 exam objectives are publicly available on the official Microsoft certification page. Always refer to the latest version of this document to ensure your study is aligned with the current exam content. Keywords to spot: 'Skills Measured', 'Exam DP-900'.
Getting Started: Navigating the AI-900 Exam
Effective Study Strategies for AI-900
Only passively reading material without active recall or practice.
Getting Started: Navigating the AI-900 Exam
Effective Study Strategies for AI-900
Ignoring the official exam objectives and studying irrelevant topics.
Getting Started: Navigating the AI-900 Exam
Effective Study Strategies for AI-900
Cramming all study into the last few days before the exam.
Getting Started: Navigating the AI-900 Exam
AI Workload
A specific task or problem that AI technologies are designed to solve.
AI Workloads and Responsible AI
Machine Learning
AI enabling systems to learn from data without explicit programming.
AI Workloads and Responsible AI
Computer Vision
AI enabling computers to interpret visual data like images and video.
AI Workloads and Responsible AI
Natural Language Processing
AI enabling computers to understand, interpret, and generate human language.
AI Workloads and Responsible AI
Knowledge Mining
AI extracting information from unstructured data to make it searchable.
AI Workloads and Responsible AI
Azure Machine Learning
Cloud platform for building, training, and deploying ML models.
AI Workloads and Responsible AI
Azure AI Language
Suite of NLP services for text analysis, translation, and more.
AI Workloads and Responsible AI
Azure AI Search
Combines search with AI for knowledge mining from diverse content.
AI Workloads and Responsible AI
Common AI Workloads & Their Features
My Computer Never Knew: Machine learning, Computer vision, Natural language processing, Knowledge mining. Remember these core workloads!
AI Workloads and Responsible AI
Common AI Workloads & Their Features
The exam frequently asks you to match a business problem or scenario to the most appropriate Azure AI workload or service. Pay close attention to keywords like 'predicting,' 'seeing,' 'understanding text,' or 'finding insights in documents.'
AI Workloads and Responsible AI
Common AI Workloads & Their Features
Confusing Computer Vision with Knowledge Mining; Computer Vision is about images, Knowledge Mining is about text.
AI Workloads and Responsible AI
Common AI Workloads & Their Features
Assuming all AI problems require custom Machine Learning models; many can be solved with pre-built cognitive services.
AI Workloads and Responsible AI
Common AI Workloads & Their Features
Not considering the type of data (text, image, numerical) when choosing an AI workload.
AI Workloads and Responsible AI
Fairness in AI
AI systems treating all individuals/groups equitably without prejudice.
AI Workloads and Responsible AI
Inclusivity in AI
Designing AI for diverse user needs, empowering all individuals.
AI Workloads and Responsible AI
Bias (AI)
Systematic error or prejudice in data or algorithms leading to unfair outcomes.
AI Workloads and Responsible AI
Data Bias
Training data not accurately reflecting real-world distribution or containing prejudices.
AI Workloads and Responsible AI
Algorithmic Bias
Bias arising from the design or optimization of the AI model itself.
AI Workloads and Responsible AI
Representational Bias
Underrepresentation or misrepresentation of certain groups in data.
AI Workloads and Responsible AI
Disparate Impact
When a neutral policy or system disproportionately harms a protected group.
AI Workloads and Responsible AI
Responsible AI: Fairness & Inclusivity
F-A-I-R: Focus on All Individuals' Rights. Remember that AI should respect everyone!
AI Workloads and Responsible AI
Responsible AI: Fairness & Inclusivity
The AI-900 exam often tests your understanding of how to identify and mitigate bias. Look for keywords like 'disproportionate impact,' 'underrepresentation,' or 'equitable outcomes' in questions related to fairness.
AI Workloads and Responsible AI
Responsible AI: Fairness & Inclusivity
Assuming that 'unbiased data' automatically leads to 'fair AI' without considering algorithmic or interaction biases.
AI Workloads and Responsible AI
Responsible AI: Fairness & Inclusivity
Focusing only on average performance metrics without evaluating performance across different demographic groups.
AI Workloads and Responsible AI
Responsible AI: Fairness & Inclusivity
Believing that AI systems are inherently neutral and cannot perpetuate or amplify human biases.
AI Workloads and Responsible AI
Reliability
AI system's consistent and correct performance over time.
AI Workloads and Responsible AI
Safety
Preventing AI systems from causing harm to people or property.
AI Workloads and Responsible AI
Human-in-the-loop
Human oversight and intervention in AI decision-making.
AI Workloads and Responsible AI
Fail-safe
System design to revert to a safe state upon failure.
AI Workloads and Responsible AI
Generalization
AI's ability to perform well on unseen data.
AI Workloads and Responsible AI
Adversarial testing
Testing AI with intentionally deceptive inputs.
AI Workloads and Responsible AI
Responsible AI: Reliability & Safety
R&S: 'R' is for 'Reliable Results', 'S' is for 'Safe Systems'. Think of a reliable car (always starts) that's also safe (has airbags).
AI Workloads and Responsible AI
Responsible AI: Reliability & Safety
The exam often tests your understanding of the 'why' behind these principles. Look for keywords like 'consistent performance' for reliability and 'preventing harm' or 'human oversight' for safety.
AI Workloads and Responsible AI
Responsible AI: Reliability & Safety
Confusing reliability (consistent performance) with safety (preventing harm).
AI Workloads and Responsible AI
Responsible AI: Reliability & Safety
Assuming AI systems are inherently safe without explicit design for safety.
AI Workloads and Responsible AI
Responsible AI: Reliability & Safety
Neglecting continuous monitoring after deployment, leading to performance degradation.
AI Workloads and Responsible AI
Differential Privacy
Adds noise to data to protect individual identities.
AI Workloads and Responsible AI
Federated Learning
Trains AI models on decentralized data sources.
AI Workloads and Responsible AI
Homomorphic Encryption
Allows computation on encrypted data without decryption.
AI Workloads and Responsible AI
Adversarial Attack
Malicious input designed to trick an AI model.
AI Workloads and Responsible AI
Data Minimization
Collecting only essential data for a purpose.
AI Workloads and Responsible AI
GDPR
EU regulation on data protection and privacy.
AI Workloads and Responsible AI
HIPAA
US law protecting patient health information.
AI Workloads and Responsible AI
Responsible AI: Privacy & Security
P.R.I.V.A.C.Y. - Protect, Restrict, Isolate, Verify, Anonymize, Control, Yield (to user consent).
AI Workloads and Responsible AI
Responsible AI: Privacy & Security
The exam often tests your understanding of privacy-preserving techniques like differential privacy and federated learning. Look for questions about how to protect sensitive data in AI systems and common regulatory frameworks like GDPR.
AI Workloads and Responsible AI
Responsible AI: Privacy & Security
Confusing privacy (data rights) with security (data protection).
AI Workloads and Responsible AI
Responsible AI: Privacy & Security
Underestimating the importance of regulatory compliance in AI projects.
AI Workloads and Responsible AI
Responsible AI: Privacy & Security
Ignoring the specific vulnerabilities of AI models to adversarial attacks.
AI Workloads and Responsible AI
Transparency
Ability to understand an AI system's workings and decisions.
AI Workloads and Responsible AI
Accountability
Framework ensuring responsibility for AI system outcomes.
AI Workloads and Responsible AI
Explainable AI (XAI)
Tools and techniques to make AI models understandable.
AI Workloads and Responsible AI
Feature Importance
Measure of how much each input feature impacts a prediction.
AI Workloads and Responsible AI
Counterfactual Explanations
Shows minimal input changes for a different AI prediction.
AI Workloads and Responsible AI
AI Governance
Policies and structures for managing AI development and use.
AI Workloads and Responsible AI
Audit Trail
Record of changes and decisions made during AI development.
AI Workloads and Responsible AI
Responsible AI: Transparency & Accountability
Imagine a 'Transparent Window' into the AI's brain, showing its decisions, and a 'Accountable Anchor' holding someone responsible for its actions. Window for Transparency, Anchor for Accountability!
AI Workloads and Responsible AI
Responsible AI: Transparency & Accountability
The exam often tests your understanding of why transparency and accountability are important for 'trust' and 'ethical' AI. Look for keywords like 'explainability,' 'interpretability,' 'responsibility,' and 'governance' in questions.
AI Workloads and Responsible AI
Responsible AI: Transparency & Accountability
Confusing transparency with simply open-sourcing code; transparency is about interpretability, not just access.
AI Workloads and Responsible AI
Responsible AI: Transparency & Accountability
Believing accountability only applies after a problem occurs; it must be built into the AI lifecycle from the start.
AI Workloads and Responsible AI
Responsible AI: Transparency & Accountability
Underestimating the legal and reputational risks of lacking clear transparency and accountability mechanisms.
AI Workloads and Responsible AI
Supervised Learning
ML type learning from labeled input-output pairs.
Machine Learning Fundamentals on Azure
Unsupervised Learning
ML type finding patterns in unlabeled data.
Machine Learning Fundamentals on Azure
Reinforcement Learning
ML type where agent learns by trial and error in an environment.
Machine Learning Fundamentals on Azure
Classification
Supervised task predicting discrete categories (e.g., spam/not spam).
Machine Learning Fundamentals on Azure
Regression
Supervised task predicting continuous values (e.g., house prices).
Machine Learning Fundamentals on Azure
Clustering
Unsupervised task grouping similar data points together.
Machine Learning Fundamentals on Azure
Labeled Data
Data with known correct outputs for each input.
Machine Learning Fundamentals on Azure
Unlabeled Data
Data without known correct outputs.
Machine Learning Fundamentals on Azure
Introduction to Machine Learning Types
Imagine 'S-U-R-F': Supervised uses a 'S'tudent, Unsupervised 'U'ncovers, Reinforcement 'R'ewards 'F'or actions.
Machine Learning Fundamentals on Azure
Introduction to Machine Learning Types
The AI-900 exam frequently presents scenarios and asks you to identify the appropriate machine learning type. Look for keywords like 'labeled data,' 'predicting a specific outcome,' 'grouping similar items,' 'finding hidden patterns,' or 'learning from rewards' to guide your answer.
Machine Learning Fundamentals on Azure
Introduction to Machine Learning Types
Confusing classification (discrete output) with regression (continuous output).
Machine Learning Fundamentals on Azure
Introduction to Machine Learning Types
Assuming all ML problems require labeled data; many don't (unsupervised).
Machine Learning Fundamentals on Azure
Introduction to Machine Learning Types
Applying reinforcement learning when a simpler supervised approach would suffice.
Machine Learning Fundamentals on Azure
Introduction to Machine Learning Types
Not considering the cost and feasibility of data labeling for supervised tasks.
Machine Learning Fundamentals on Azure
Data
Raw facts and figures used to train ML models.
Machine Learning Fundamentals on Azure
Model
A program that has learned patterns from data to make predictions.
Machine Learning Fundamentals on Azure
Training
The process of teaching a model using data to adjust its parameters.
Machine Learning Fundamentals on Azure
Feature Engineering
Creating new input variables from existing ones to improve model performance.
Machine Learning Fundamentals on Azure
Overfitting
When a model learns training data too well, performing poorly on new data.
Machine Learning Fundamentals on Azure
Training Data
The dataset used to teach the machine learning model.
Machine Learning Fundamentals on Azure
Validation Data
Used for tuning model hyperparameters and preventing overfitting.
Machine Learning Fundamentals on Azure
Test Data
Used for final, unbiased evaluation of a trained model's performance.
Machine Learning Fundamentals on Azure
Core ML Concepts: Data, Models, & Training
To remember the order: 'DMT' – Data makes the Model Trained. Or for the splits: 'TVT' – Train, Validate, Test. (Like watching TV, then taking a Test!)
Machine Learning Fundamentals on Azure
Core ML Concepts: Data, Models, & Training
The exam frequently tests the distinct purposes of training, validation, and test datasets. Remember: training builds the model, validation fine-tunes it, and testing evaluates its final, unseen performance.
Machine Learning Fundamentals on Azure
Core ML Concepts: Data, Models, & Training
Using the test set for hyperparameter tuning, leading to an overly optimistic performance estimate.
Machine Learning Fundamentals on Azure
Core ML Concepts: Data, Models, & Training
Neglecting data cleaning and preprocessing, resulting in 'garbage in, garbage out' model performance.
Machine Learning Fundamentals on Azure
Core ML Concepts: Data, Models, & Training
Training a model on too little data, which can lead to poor generalization.
Machine Learning Fundamentals on Azure
Evaluation Metrics
Quantitative measures of a model's performance.
Machine Learning Fundamentals on Azure
Accuracy
Proportion of correct predictions over total predictions.
Machine Learning Fundamentals on Azure
Precision
Proportion of true positive predictions among all positive predictions.
Machine Learning Fundamentals on Azure
Recall
Proportion of true positive predictions among all actual positives.
Machine Learning Fundamentals on Azure
F1-score
Harmonic mean of precision and recall.
Machine Learning Fundamentals on Azure
Deployment
Making a trained model available for use in applications.
Machine Learning Fundamentals on Azure
Real-time Inference
Generating predictions instantly for individual requests.
Machine Learning Fundamentals on Azure
Batch Inference
Generating predictions for large datasets at once.
Machine Learning Fundamentals on Azure
Core ML Concepts: Evaluation & Deployment
To remember evaluation metrics: 'PR-F1' for classification (Precision, Recall, F1-score) and 'MAE-MSE-R2' for regression (Mean Absolute Error, Mean Squared Error, R-squared).
Machine Learning Fundamentals on Azure
Core ML Concepts: Evaluation & Deployment
The AI-900 exam expects you to know common evaluation metrics for classification (accuracy, precision, recall, F1-score) and regression (MAE, MSE, R-squared) and understand the difference between real-time and batch inference.
Machine Learning Fundamentals on Azure
Core ML Concepts: Evaluation & Deployment
Only using accuracy for imbalanced classification datasets, leading to misleading performance estimates.
Machine Learning Fundamentals on Azure
Core ML Concepts: Evaluation & Deployment
Deploying a model without proper monitoring, missing performance degradation over time.
Machine Learning Fundamentals on Azure
Core ML Concepts: Evaluation & Deployment
Confusing real-time inference with batch inference and choosing the wrong method for the use case.
Machine Learning Fundamentals on Azure
Azure ML Workspace
Centralized hub for managing all ML assets and activities in Azure.
Machine Learning Fundamentals on Azure
Compute Target
Specialized computing resource for running ML code (training or inference).
Machine Learning Fundamentals on Azure
Datastore
Reference to a storage location in Azure, like Blob Storage.
Machine Learning Fundamentals on Azure
Dataset
Versioned data within a datastore used for ML training and evaluation.
Machine Learning Fundamentals on Azure
Endpoint
A deployed model exposed as a web service for making predictions.
Machine Learning Fundamentals on Azure
Experiment
A run of ML code, tracking metrics, parameters, and outputs.
Machine Learning Fundamentals on Azure
Pipeline
A sequence of ML steps, orchestrating data prep, training, and deployment.
Machine Learning Fundamentals on Azure
Common ML Components & Tools in Azure
WCDME: **W**orkspace, **C**ompute, **D**ataset, **M**odel, **E**ndpoint. Remember these five core components as the 'Wicked Cool Data Model Empire' of Azure ML!
Machine Learning Fundamentals on Azure
Common ML Components & Tools in Azure
On the AI-900 exam, be able to identify the purpose of an Azure Machine Learning workspace, compute targets (training vs. inference), datasets, and models/endpoints. Keywords like 'centralized resource' for workspace, 'run code' for compute, 'data versioning' for datasets, and 'web service' for endpoints are important.
Machine Learning Fundamentals on Azure
Common ML Components & Tools in Azure
Confusing a datastore with a dataset: A datastore is the *location* of data, while a dataset is a *versioned view* of that data.
Machine Learning Fundamentals on Azure
Common ML Components & Tools in Azure
Using a training compute target for high-scale real-time inference: Training targets are optimized for large, iterative computations, not low-latency, high-throughput serving.
Machine Learning Fundamentals on Azure
Common ML Components & Tools in Azure
Not registering models: Skipping model registration loses versioning and metadata, making reproducibility and tracking difficult.
Machine Learning Fundamentals on Azure
Azure ML Studio
Web-based IDE for the entire ML lifecycle, offering visual and code-first tools.
Machine Learning Fundamentals on Azure
Automated ML (AutoML)
Feature in Azure ML Studio that automates model selection, training, and tuning.
Machine Learning Fundamentals on Azure
Workspace
A centralized place in Azure ML Studio to manage ML assets and projects.
Machine Learning Fundamentals on Azure
Hyperparameters
Configuration settings for a machine learning algorithm, tuned during training.
Machine Learning Fundamentals on Azure
Model Deployment
Making a trained ML model available for use by other applications, often as an endpoint.
Machine Learning Fundamentals on Azure
Azure ML Studio & Automated ML Overview
Think 'AutoML is your Auto-Pilot for Model building!' It takes over the tedious parts so you can focus on the destination.
Machine Learning Fundamentals on Azure
Azure ML Studio & Automated ML Overview
The exam often asks about the primary benefit of AutoML, which is its ability to automate model selection and hyperparameter tuning, leading to faster model development and reduced need for deep ML expertise. Remember it's a feature within Azure ML Studio, not a standalone service.
Machine Learning Fundamentals on Azure
Azure ML Studio & Automated ML Overview
Confusing Azure ML Studio as only a visual tool; it also supports extensive code-first development.
Machine Learning Fundamentals on Azure
Azure ML Studio & Automated ML Overview
Believing AutoML replaces the entire ML lifecycle; it primarily automates model training and tuning, not data preparation or deployment.
Machine Learning Fundamentals on Azure
Azure ML Studio & Automated ML Overview
Assuming AutoML always finds the 'best' model for every scenario; it finds a high-performing model based on specified metrics and data, but custom solutions might still outperform it.
Machine Learning Fundamentals on Azure
Image Analysis
Extracting descriptive information and context from images.
Computer Vision on Azure
Scene Understanding
AI's ability to describe the overall environment in an image.
Computer Vision on Azure
Content Moderation
Automated identification of inappropriate visual content.
Computer Vision on Azure
Spatial Reasoning
Understanding object locations and relationships in space.
Computer Vision on Azure
Azure AI Vision
Azure service for pre-trained computer vision models.
Computer Vision on Azure
Azure Custom Vision
Azure service for building custom computer vision models.
Computer Vision on Azure
Common Computer Vision Capabilities
To remember the breadth of Computer Vision: 'C'an 'V'iew 'I'mages 'O'r 'N'ot? (CVION: Classification, Vision, Image Analysis, Object Detection, OCR, N-dimensional data).
Computer Vision on Azure
Common Computer Vision Capabilities
The AI-900 exam tests your understanding of what computer vision is and its common capabilities. Focus on differentiating between tasks like image classification, object detection, and general image analysis. Keywords to spot include 'interpret visual data', 'understand content', and specific applications like 'quality control' or 'medical image analysis'.
Computer Vision on Azure
Common Computer Vision Capabilities
Confusing image classification (what is in the image?) with object detection (where are the specific objects in the image?).
Computer Vision on Azure
Common Computer Vision Capabilities
Underestimating the breadth of computer vision; it's not just about identifying objects but also understanding context and relationships.
Computer Vision on Azure
Common Computer Vision Capabilities
Assuming all computer vision tasks require custom model training; many common tasks can use pre-trained Azure services.
Computer Vision on Azure
Image Classification
Assigning a single category label to an entire image.
Computer Vision on Azure
Object Detection
Identifying and locating multiple objects within an image with bounding boxes.
Computer Vision on Azure
Bounding Box
A rectangular coordinate that outlines an object in an image.
Computer Vision on Azure
Convolutional Neural Network (CNN)
A deep learning model specialized for processing image data.
Computer Vision on Azure
Azure Computer Vision API
Azure service providing pre-trained computer vision capabilities.
Computer Vision on Azure
Image Classification & Object Detection
Classify is 'one label for the whole class'. Detect is 'find many objects and draw boxes'.
Computer Vision on Azure
Image Classification & Object Detection
The exam often tests your ability to distinguish between image classification and object detection. Look for keywords like 'single label for the whole image' (classification) versus 'identifying and locating multiple items' (detection).
Computer Vision on Azure
Image Classification & Object Detection
Confusing image classification with object detection; classification gives one overall label, detection finds many specific items.
Computer Vision on Azure
Image Classification & Object Detection
Assuming object detection is always needed; if you only care about the general theme, classification is simpler and faster.
Computer Vision on Azure
Image Classification & Object Detection
Forgetting that both can be implemented using Azure services like Custom Vision or Computer Vision API.
Computer Vision on Azure
Facial Recognition
Identifies individuals by analyzing unique facial features in images or video.
Computer Vision on Azure
Faceprint
A mathematical representation of unique facial features used for identification.
Computer Vision on Azure
Face Detection
Identifies the presence and location of human faces in an image or video.
Computer Vision on Azure
Optical Character Recognition (OCR)
Converts images of text (scanned documents, photos) into editable digital text.
Computer Vision on Azure
Preprocessing
Initial steps in OCR to clean and enhance an image before character recognition.
Computer Vision on Azure
Character Segmentation
The process of isolating individual characters within an image for OCR.
Computer Vision on Azure
Facial Recognition & Optical Character Recognition (OCR)
To remember OCR: 'O' for 'Optical' (seeing the text), 'C' for 'Character' (individual letters), 'R' for 'Recognition' (understanding them).
Computer Vision on Azure
Facial Recognition & Optical Character Recognition (OCR)
For the AI-900 exam, remember that facial recognition focuses on *identifying* individuals, while OCR is about converting *images of text* into searchable, editable digital text. Look for keywords like 'identity verification' for facial recognition and 'extract text from image' for OCR.
Computer Vision on Azure
Facial Recognition & Optical Character Recognition (OCR)
Confusing face detection (finding a face) with facial recognition (identifying a person).
Computer Vision on Azure
Facial Recognition & Optical Character Recognition (OCR)
Underestimating the ethical and privacy implications of facial recognition systems.
Computer Vision on Azure
Facial Recognition & Optical Character Recognition (OCR)
Not realizing that OCR can handle both printed and handwritten text with varying degrees of accuracy.
Computer Vision on Azure
Azure Computer Vision
Pre-trained API for general image analysis, OCR, and object detection.
Computer Vision on Azure
Azure Face Service
Dedicated API for detecting, analyzing, and identifying human faces.
Computer Vision on Azure
API Key
A secret code used to authenticate and authorize access to an API.
Computer Vision on Azure
SDK (Software Development Kit)
Tools and libraries that simplify interaction with an API in a programming language.
Computer Vision on Azure
REST API
A set of rules allowing web services to communicate, often returning JSON.
Computer Vision on Azure
Model Training
The process of teaching a machine learning model using labeled data.
Computer Vision on Azure
Azure Computer Vision Tools & Services
Imagine a 'C' for Computer Vision (general), a 'C' for Custom Vision (specific), and an 'F' for Face (faces). C-C-F: Common, Custom, Face!
Computer Vision on Azure
Azure Computer Vision Tools & Services
The exam often distinguishes between general-purpose Computer Vision and specialized services like Custom Vision or Face. Remember: if it's general image analysis, think Azure Computer Vision; if it's unique objects or classifications, think Custom Vision; if it's human faces, think Face Service.
Computer Vision on Azure
Azure Computer Vision Tools & Services
Confusing Azure Computer Vision (general) with Azure Custom Vision (specific, custom-trained).
Computer Vision on Azure
Azure Computer Vision Tools & Services
Not understanding that the Face Service is distinct from the general Computer Vision service.
Computer Vision on Azure
Azure Computer Vision Tools & Services
Forgetting that API keys are essential for authenticating requests to Azure AI services.
Computer Vision on Azure
Natural Language Processing (NLP)
AI branch enabling computers to understand and process human language.
Natural Language Processing on Azure
Named Entity Recognition (NER)
Identifies and classifies key entities like people, places, organizations in text.
Natural Language Processing on Azure
Text Summarization
Condenses long texts into shorter, coherent summaries.
Natural Language Processing on Azure
Extractive Summarization
Creates summaries by pulling key sentences directly from the original text.
Natural Language Processing on Azure
Abstractive Summarization
Generates new sentences to capture the main ideas of a text.
Natural Language Processing on Azure
Language Detection
Identifies the natural language in which a piece of text is written.
Natural Language Processing on Azure
Machine Translation
Automatically converts text or speech from one language to another.
Natural Language Processing on Azure
Common NLP Workload Capabilities
To remember common NLP capabilities, think 'ELASTIC': **E**ntity Recognition, **L**anguage Detection, **A**nalysis (Sentiment), **S**ummarization, **T**ranslation, **I**nterpretation (LUI/LUIS), **C**ontent Moderation.
Natural Language Processing on Azure
Common NLP Workload Capabilities
The exam often tests your ability to match an NLP task to its real-world application. For example, if a scenario describes extracting names and locations, think 'Named Entity Recognition.' If it's about condensing a long document, think 'Text Summarization.'
Natural Language Processing on Azure
Common NLP Workload Capabilities
Confusing extractive summarization with abstractive summarization; remember, extractive uses original sentences.
Natural Language Processing on Azure
Common NLP Workload Capabilities
Underestimating the importance of language detection as a prerequisite for many other NLP tasks.
Natural Language Processing on Azure
Common NLP Workload Capabilities
Thinking NLP is only about understanding; it also includes generating and moderating language.
Natural Language Processing on Azure
Key Phrase Extraction
Identifies the most important topics or concepts in a text.
Natural Language Processing on Azure
Sentiment Analysis
Determines the emotional tone (positive, negative, neutral) of text.
Natural Language Processing on Azure
Opinion Mining
Another term for sentiment analysis, focusing on public opinion.
Natural Language Processing on Azure
Azure Language Service
A cloud-based AI service offering NLP capabilities like sentiment analysis.
Natural Language Processing on Azure
Unstructured Data
Information that does not have a predefined data model, like text.
Natural Language Processing on Azure
Confidence Score
A numerical value indicating the certainty of a model's prediction.
Natural Language Processing on Azure
Key Phrase Extraction & Sentiment Analysis
KPE (Key Phrase Extraction) finds the 'Key' topics. SA (Sentiment Analysis) tells you the 'Sentiment' or feeling. Keys unlock topics, feelings are sentiments!
Natural Language Processing on Azure
Key Phrase Extraction & Sentiment Analysis
The AI-900 exam expects you to differentiate between key phrase extraction (what is being talked about) and sentiment analysis (how people feel about it). Remember that Azure's Language service provides these capabilities as pre-built AI services.
Natural Language Processing on Azure
Key Phrase Extraction & Sentiment Analysis
Confusing key phrase extraction with summarization; KPE identifies topics, summarization condenses the entire text.
Natural Language Processing on Azure
Key Phrase Extraction & Sentiment Analysis
Assuming sentiment analysis can understand complex sarcasm or irony without advanced custom training.
Natural Language Processing on Azure
Key Phrase Extraction & Sentiment Analysis
Trying to build these NLP models from scratch when Azure offers powerful, pre-trained services.
Natural Language Processing on Azure
Language Understanding (LUIS)
Azure service to interpret natural language, identifying intents and entities.
Natural Language Processing on Azure
Intent
The goal or purpose expressed in a user's natural language utterance.
Natural Language Processing on Azure
Entity
Specific, relevant pieces of information extracted from a user's utterance.
Natural Language Processing on Azure
Speech-to-Text (STT)
Converts spoken audio into written text.
Natural Language Processing on Azure
Text-to-Speech (TTS)
Converts written text into synthesized, human-like spoken audio.
Natural Language Processing on Azure
Utterance
A single spoken or typed input from a user to a conversational AI system.
Natural Language Processing on Azure
Neural Voices
Highly natural and expressive synthesized voices generated using deep learning.
Natural Language Processing on Azure
Language Understanding & Speech Recognition
L-U-I-S: 'L'anguage 'U'nderstands 'I'ntents & 'S'tuff. Speech-to-Text is 'S'peaking to 'T'exting; Text-to-Speech is 'T'exting to 'S'peaking.
Natural Language Processing on Azure
Language Understanding & Speech Recognition
For the AI-900 exam, remember that LUIS is specifically for understanding the *meaning* (intents and entities) of language, while Speech-to-Text and Text-to-Speech handle the *conversion* between audio and text. Understand their distinct roles and how they integrate.
Natural Language Processing on Azure
Language Understanding & Speech Recognition
Confusing Speech-to-Text with LUIS: STT transcribes words; LUIS understands their meaning.
Natural Language Processing on Azure
Language Understanding & Speech Recognition
Underestimating the importance of training data for LUIS: Poorly trained LUIS models won't accurately identify intents and entities.
Natural Language Processing on Azure
Language Understanding & Speech Recognition
Not considering the need for custom speech models for specialized vocabulary: Generic speech models might struggle with industry-specific jargon.
Natural Language Processing on Azure
Azure AI Speech
Service for converting spoken language to text and text to speech.
Natural Language Processing on Azure
Speech-to-Text
Transcribes audio input into written text.
Natural Language Processing on Azure
Text-to-Speech
Converts written text into natural-sounding spoken audio.
Natural Language Processing on Azure
Azure AI Translator
Cloud service for real-time, neural machine translation across languages.
Natural Language Processing on Azure
Azure NLP Tools & Services
LST: Language for Text, Speech for Talk, Translator for Tongues. It helps you remember which service does what!
Natural Language Processing on Azure
Azure NLP Tools & Services
Memorize the core purpose of each service: Azure AI Language for text insights, Azure AI Speech for voice, and Azure AI Translator for multilingual text. The exam often tests your ability to choose the right service for a given scenario.
Natural Language Processing on Azure
Azure NLP Tools & Services
Confusing Azure AI Language with Azure AI Speech: Language is for text processing, Speech is for audio processing.
Natural Language Processing on Azure
Azure NLP Tools & Services
Trying to perform translation using Azure AI Language instead of Azure AI Translator.
Natural Language Processing on Azure
Azure NLP Tools & Services
Overlooking the pre-trained models available, attempting to build custom models for common tasks unnecessarily.
Natural Language Processing on Azure