Update 5-exercise.md

Author: kprks

learn-pr/philanthropies/forging-voices-from-data/includes/5-exercise.md

@@ -1,71 +1,59 @@
-### Hands-On Activity: Train and Customize a Dialogue Model in Azure OpenAI Studio  
-**Time:** 20–25 min
-
----
-
 **Objective:**  
-Enhance the chatbot created in Module 1 by integrating a custom-trained dialogue model using Azure OpenAI Studio. Use domain-specific data to improve response accuracy, refine tone, and simulate memory through prompt engineering. Test and iterate on the chatbot’s performance in real time.
 
----
+Enhance the chatbot created in module 1 by integrating a custom-trained dialogue model using Azure OpenAI Studio. Use domain-specific data to improve response accuracy, refine tone, and simulate memory through prompt engineering. Test and iterate on the chatbot's performance in real time.
+
+**Materials needed:** 
 
-**Materials Needed:**  
 - Laptop or desktop with internet access  
 - Azure OpenAI Studio access  
 - Dataset of sample dialogues or FAQs (provided or created by learner)
 
----
+## Part 1: Upload and train with custom data
 
-### Part 1: Upload and Train with Custom Data
-
-1. **Prepare a Dataset**  
+1. **Prepare a dataset**  
    - Use a collection of relevant user queries and ideal responses (such as customer FAQs, support logs, or product descriptions).  
    - Format the data in CSV or plain text, with clear input/output pairs.
 
-2. **Upload the Dataset**  
+2. **Upload the dataset**  
    - In Azure OpenAI Studio, navigate to your chatbot project.  
    - Upload your dataset under the **Files** section and connect it to your deployment using the Playground.
 
-3. **Apply Supervised Learning via Prompt Examples**  
+3. **Apply supervised learning via prompt examples**  
    - Provide a few sample dialogues using system + user + assistant messages to demonstrate ideal behavior.  
      Example:
      ```
      User: "How do I reset my password?"  
      Assistant: "To reset your password, go to the login page and click 'Forgot Password.' Follow the instructions sent to your email."
      ```
 
----
-
-### Part 2: Customize Tone and Interaction Style
+## Part 2: Customize tone and interaction style
 
-1. **Set a System Prompt**  
-   - Define the assistant’s role and tone.  
+1. **Set a system prompt**  
+   - Define the assistant's role and tone.  
      Example:  
-     *“You are a friendly and professional assistant for an online bookstore. Respond with clear, concise, and warm answers.”*
+     *"You are a friendly and professional assistant for an online bookstore. Respond with clear, concise, and warm answers."*
 
-2. **Simulate Context Awareness**  
+2. **Simulate context awareness**  
    - Create a prompt that includes conversation history, enabling the model to handle follow-up questions.  
      Example:  
-     *“The user previously asked about shipping policies. Now they are asking, ‘How long does express shipping take?’”*
+     *"The user previously asked about shipping policies. Now they are asking, 'How long does express shipping take?'"*
 
-3. **Test and Iterate**  
+3. **Test and iterate**  
    - Interact with your chatbot in the Playground.  
    - Ask questions covered in your dataset and test how it handles follow-ups.  
    - Tweak your examples or prompt instructions to improve continuity and engagement.
 
----
-
-### Part 3: Evaluate and Improve
+## Part 3: Evaluate and improve
 
-1. **Identify Strengths and Gaps**  
+1. **Identify strengths and gaps**  
    - Does the chatbot handle follow-ups appropriately?  
    - Does it reflect the desired tone and style?  
    - Are any answers too vague, repetitive, or off-topic?
 
-2. **Refine and Re-Test**  
+2. **Refine and re-test**  
    - Update the dataset, adjust the prompt, or provide additional sample dialogues.  
    - Test variations in the Playground until the responses align with expectations.
 
----
+## Expected outcome
 
-**Expected Outcome:**  
 Learners will transform a basic chatbot into a refined, domain-specific assistant by uploading a custom dataset, configuring system prompts, and simulating context awareness. This activity demonstrates the power of Azure OpenAI Studio for training and fine-tuning dialogue behavior to better meet user needs and organizational goals.