Azure AI Language Service

Quick Answer: Azure AI Language is the unified NLP service providing both pre-configured features (sentiment analysis, NER, key phrases, PII detection) and customizable features (CLU for intent/entity extraction, custom question answering, custom text classification). It replaces several older services including LUIS and QnA Maker.

What Is Azure AI Language?

Azure AI Language is Microsoft's comprehensive NLP service that provides a wide range of text analysis capabilities. It consolidates multiple NLP features under a single service with a unified endpoint, making it the go-to service for text-related AI tasks on Azure.

Pre-Configured Features (No Training Required)

These features work immediately — send text and get results:

Customizable Features (Training Required)

These features require training with your data:

Conversational Language Understanding (CLU)

CLU enables you to build models that understand the intent (what the user wants to do) and extract entities (key information) from natural language utterances.

CLU Replaces LUIS

CLU is the successor to LUIS (Language Understanding Intelligent Service), which was retired in October 2025. If you see references to LUIS on the exam, know that CLU is the current replacement.

CLU Concepts

How CLU Works

1. Define intents — the actions users want to perform (BookFlight, CancelFlight, CheckStatus)
2. Define entities — the key information to extract (destination, date, passenger name)
3. Add utterances — provide example sentences for each intent with entities labeled
4. Train the model — CLU learns patterns from your labeled examples
5. Test and publish — test with new utterances and deploy the model
6. Use in applications — send user input to CLU and receive intent + entities

CLU Example

Training utterances for BookFlight intent:

• "Book a flight to Paris next Friday" → Intent: BookFlight, Entities: {destination: Paris, date: next Friday}
• "I need to fly to London tomorrow" → Intent: BookFlight, Entities: {destination: London, date: tomorrow}
• "Reserve a seat to Tokyo on March 15" → Intent: BookFlight, Entities: {destination: Tokyo, date: March 15}

After training, a new utterance:

• "Find me a flight to Berlin next week" → Intent: BookFlight (0.95), Entities: {destination: Berlin, date: next week}

On the Exam: CLU is frequently tested. Know that it identifies INTENTS (what the user wants) and ENTITIES (key details) from natural language. It requires training with labeled utterances. LUIS is retired — CLU is the replacement.

Custom Question Answering

Custom question answering creates a knowledge base that answers natural language questions. It builds on your existing content:

How It Works

1. Add sources — FAQ pages, product manuals, documents, or manually entered Q&A pairs
2. Build the knowledge base — the service extracts question-answer pairs automatically
3. Test — ask questions and verify answers
4. Publish — deploy as an endpoint for chatbots and applications

Question Answering Sources

Key Features

• Multi-turn conversations — follow-up questions that reference previous context
• Chit-chat — built-in responses for casual conversation (greetings, jokes)
• Synonyms — understand different phrasings of the same question
• Confidence scores — each answer includes a confidence rating

On the Exam: Custom question answering replaces QnA Maker (retired). It is for building FAQ-style chatbots that answer questions from your existing documents. CLU is for understanding intent and entities; question answering is for finding specific answers in your knowledge base.

When to Use Which Feature