SEO Title: Day 6: Merging Emotion + GPT Analysis into One Score #trustscore #fusionai
Focus Keyphrase: AI lie detector
Meta Description: Learn how to create a truth probability score in your AI lie detector app by combining GPT-based text analysis and emotion-based voice signals.
Introducing the Truth Score™
We now have two powerful signals in our AI lie detector:
- 🎤 Vocal emotion score from hesitation detection
- 🧠 Semantic contradiction analysis from GPT
Let’s combine them into a single, user-friendly **Truth Score** between 0–100 to represent the confidence in the truthfulness of a statement.
Step 1: Assign Numeric Scores to GPT Output
def gpt_score_label(gpt_result):
if "Likely Truthful" in gpt_result:
return 90
elif "Possibly Uncertain" in gpt_result:
return 60
elif "Suspicious" in gpt_result:
return 30
else:
return 50 # fallback
Step 2: Normalize the Emotion Score
We assume more pauses = more nervousness. You can normalize based on observed range (e.g. 0–20):
def emotion_score_from_pauses(pause_count):
score = max(0, 100 - pause_count * 5)
return min(score, 100)
Step 3: Merge into a Weighted Truth Score
def final_truth_score(gpt_score, emotion_score):
# You can tweak weights here!
return round((gpt_score * 0.6) + (emotion_score * 0.4))
Step 4: Display the Score in the API Response
@app.route('/api/voice', methods=['POST'])
def receive_audio():
...
gpt_result = analyze_with_gpt(transcript["text"])
gpt_score = gpt_score_label(gpt_result)
emotion_score = emotion_score_from_pauses(pause_score)
truth_score = final_truth_score(gpt_score, emotion_score)
return {
"transcript": transcript["text"],
"gpt_analysis": gpt_result,
"emotion_score": emotion_score,
"truth_score": truth_score
}
Step 5: Frontend Preview (Tomorrow)
Tomorrow we’ll build a visual frontend to show:
- 📝 The transcript
- 🧠 GPT label
- 🎧 Emotion score
- ✅ Final Truth Score (as a colored progress bar)
Why Truth Scoring Is Crucial
Now your AI lie detector delivers something tangible — a single score that reflects both how something was said and what was said.
Coming Up in Day 7
In Day 7: Building a Frontend Truth Meter UI #progressbar #emotionUX, we’ll bring this score to life with a live progress bar and semantic indicators.
Tags: #AIUX #TruthScore #LieDetection #MultimodalAI
