1. What This Means in the Exam
Using Azure AI Foundry, you should understand how to:
Use text-based AI (language models, text analytics)
Use speech-based AI (speech recognition and synthesis)
Combine them into real-world solutions
Deploy and test them
👉 Think:
“How do I build apps that understand/generate text and speech?”
📝 2. Implementing AI for Text
📌 What “Text AI” Includes
🔤 Text Analysis
Sentiment analysis
Entity recognition
Key phrase extraction
✍️ Generative Text
Chatbots
Summarisation
Content generation
⚙️ How It’s Implemented in Foundry
Select a language model (LLM)
Create prompts (system + user)
(Optional) Add grounding data
Test in playground
Deploy as API/app
✅ Example Scenario
“Build a chatbot that answers customer queries using company documentation”
✔️ Use:
LLM (text model)
Prompt engineering
Grounding
🗣️ 3. Implementing AI for Speech
📌 What “Speech AI” Includes
🎤 Speech Recognition
Audio → Text
🔊 Speech Synthesis
Text → Audio
⚙️ How It’s Implemented in Foundry
Select a speech model
Configure:
Language
Voice (for synthesis)
Send audio/text input
Receive transcription or spoken output
✅ Example Scenarios
Scenario 1
“Convert customer calls into transcripts”
✔️ Use:
Speech recognition
Scenario 2
“Read out notifications to users”
✔️ Use:
Speech synthesis
🔄 4. Combining Text + Speech (Very Important)
Many solutions combine both:
🎯 Example: Voice Assistant
Flow:
User speaks → Speech recognition
Text processed → Text model (LLM)
Response generated → Text
Response spoken → Speech synthesis
⚖️ 5. How to Identify the Correct Approach (Exam Logic)
Requirement AI Capability
Understand text meaning Text analysis
Generate text Generative AI
Convert speech to text Speech recognition
Convert text to speech Speech synthesis
Voice-based interaction Combine both
⚠️ 6. Common Exam Traps
❌ Confusing text analysis vs generation
✔️ Analysis = understand
✔️ Generation = create
❌ Mixing up speech directions
✔️ Audio → Text = Recognition
✔️ Text → Audio = Synthesis
❌ Ignoring combined scenarios
✔️ Many questions require both text + speech
🧠 7. Simple Memory Model
“Read, Write, Hear, Speak”
Read → Text analysis
Write → Text generation
Hear → Speech recognition
Speak → Speech synthesis
✅ Summary
To implement AI solutions for text and speech using Foundry:
Use Azure AI Foundry
For text:
Use LLMs and text analytics
Apply prompts and grounding
For speech:
Use speech recognition and synthesis
Combine them for voice-enabled AI apps