Update app.py

app.py

@@ -12,12 +12,15 @@ def split_sentences(paragraph):
     """Split a paragraph into sentences."""
     return re.split(r'(?<=[.!?]) +', paragraph.strip())
 
+def group_sentences(sentences, size=2):
+    """Group sentences into chunks of 2 (or remaining)."""
+    return [" ".join(sentences[i:i+size]) for i in range(0, len(sentences), size)]
+
 def detect_ai(text):
-    # Split into paragraphs
     paragraphs = re.split(r"\n\s*\n", text.strip())
 
     results = []
-    all_scores = []
+    all_ai_flags = []  # store 1 = AI, 0 = Human
 
     highlighted = ""
     for para in paragraphs:
@@ -25,50 +28,57 @@ def detect_ai(text):
             continue
 
         sentences = split_sentences(para)
+        chunks = group_sentences(sentences, size=2)
         highlighted_para = ""
 
-        for sent in sentences:
-            if not sent.strip():
+        for chunk in chunks:
+            if not chunk.strip():
                 continue
 
-            inputs = tokenizer(sent, return_tensors="pt", truncation=True, max_length=512)
+            inputs = tokenizer(chunk, return_tensors="pt", truncation=True, max_length=512)
             with torch.no_grad():
                 outputs = model(**inputs)
                 probs = torch.softmax(outputs.logits, dim=1)
 
             ai_score = float(probs[0][1])   # AI likelihood
             human_score = 1 - ai_score      # Human likelihood
-            all_scores.append(human_score)
 
-            # Decide label
-            if human_score < 0.9:
+            # Threshold check (AI > 20% → AI)
+            if ai_score > 0.2:
                 label = "🔴 AI"
                 color = "rgb(255,120,120)"  # red
+                all_ai_flags.append(1)
             else:
                 label = "🟢 Human"
                 color = "rgb(120,255,120)"  # green
+                all_ai_flags.append(0)
 
             highlighted_para += (
-                f"<span style='background-color:{color}; padding:2px; border-radius:3px' "
-                f"title='{label} — Human {round(human_score*100,1)}% | AI {round(ai_score*100,1)}%'>"
-                f"{sent} </span>"
+                f"<div style='background-color:{color}; padding:4px; margin-bottom:4px; border-radius:4px'>"
+                f"<b>{label}</b> — Human {round(human_score*100,1)}% | AI {round(ai_score*100,1)}%<br>"
+                f"{chunk}</div>"
             )
 
-        highlighted += f"<div style='margin-bottom:10px'>{highlighted_para}</div>"
-
-    # Compute overall human %
-    if all_scores:
-        avg_human = sum(all_scores) / len(all_scores)
-        total_human = round(avg_human * 100, 2)
-        highlighted += f"<p><b>⚖️ Overall Human Probability: {total_human}%</b></p>"
+        highlighted += f"<div style='margin-bottom:12px'>{highlighted_para}</div>"
+
+    # Compute overall result
+    if all_ai_flags:
+        ai_ratio = sum(all_ai_flags) / len(all_ai_flags)
+        if ai_ratio == 1:
+            overall = "🔴 Overall: 100% AI"
+        elif ai_ratio == 0:
+            overall = "🟢 Overall: 100% Human"
+        else:
+            overall = f"⚖️ Overall AI Probability: {round(ai_ratio*100,2)}%"
+        highlighted += f"<p><b>{overall}</b></p>"
     else:
-        total_human = 0.0
+        overall = "No text detected"
 
-    return highlighted, {"overall_human_percent": total_human}
+    return highlighted, {"overall": overall, "chunks_checked": len(all_ai_flags)}
 
 with gr.Blocks() as demo:
-    gr.Markdown("## 🤖 AI Detector (Sentence-level)")
-    gr.Markdown("Each sentence is checked. If Human <90%, it’s flagged as AI.")
+    gr.Markdown("## 🤖 AI Detector (2-sentence chunks)")
+    gr.Markdown("Groups of 2 sentences are checked. If AI >20%, the group is flagged as AI.")
     input_text = gr.Textbox(lines=12, placeholder="Paste your essay or report here...")
     output_html = gr.HTML()
     output_json = gr.JSON()