DEMO.md

🎥 CleanSpeak Demo Guide

Quick Demo Script

Scenario 1: Non-Toxic Comment

Input:

I love this product! It works amazingly well and has helped me tremendously.

Expected Output:

✅ Toxicity Status: No

All toxicity scores should be below 50%

---

Scenario 2: Toxic Comment (Insults)

Input:

You are such a stupid idiot. This is the worst thing I've ever seen!

Expected Output:

🚨 Toxicity Detected: Yes - ☠️ Toxic, 👊 Insult

- Toxic: High score (70-90%)
- Insult: High score (70-90%)
- Other categories: Lower scores
- Toxic Words Highlighted: "stupid", "idiot"

---

Scenario 3: Severe Toxicity

Input:

I hate you so much! You should just die already. Worthless piece of garbage.

Expected Output:

🚨 Toxicity Detected: Yes - ☠️ Toxic, 💀 Severe Toxic, 👊 Insult

- Toxic: Very high (80-95%)
- Severe Toxic: High (60-80%)
- Insult: High (70-85%)
- Toxic Words Highlighted: "hate", "die", "worthless"

---

Scenario 4: Threatening Language

Input:

Watch your back. I'm coming for you.

Expected Output:

🚨 Toxicity Detected: Yes - ⚠️ Threat

- Threat: Moderate to high (50-70%)
- Other categories: Lower scores

---

Scenario 5: Obscene Content

Input:

What the f*** is wrong with you? This is complete bull***t!

Expected Output:

🚨 Toxicity Detected: Yes - ☠️ Toxic, 🔞 Obscene

- Toxic: High (60-80%)
- Obscene: High (70-85%)
- Toxic Words Highlighted: (may vary based on filters)

---

Scenario 6: Identity Hate

Input:

All [group] people are terrible and should not be allowed anywhere.

Expected Output:

🚨 Toxicity Detected: Yes - ☠️ Toxic, 🚫 Identity Hate

- Toxic: Moderate (50-70%)
- Identity Hate: Moderate to high (40-70%)
- Threat: May have some score

---

🎯 Testing Tips

1. Start with clean examples to see the "No" result
2. Gradually increase toxicity to see thresholds activate
3. Try different combinations of toxicity types
4. Test edge cases like sarcasm or context-dependent language
5. Check word highlighting to see which words trigger detection

📊 Understanding the Results

• Yes: At least one toxicity type scored ≥ 50%
• No: All toxicity types scored < 50%
• Progress bars: Visual representation of confidence scores
• Toxic Words: Keywords that likely contributed to the detection

🎨 UI Elements to Highlight

1. Gradient background: Shows on load
2. Animated header: Fade-in effect
3. Color-coded results: Red for Yes, Green for No
4. Severity bars: Gradient from red to yellow to blue
5. Toxic word highlighting: Red background with bold white text
6. Tip box: Appears when toxicity is detected