import gradio as gr from transformers import pipeline import time # ========================================== # MODEL CONFIGURATION # ========================================== MODEL_NAME = "ENTUM-AI/roberta-clickbait-classifier" print(f"Loading model: {MODEL_NAME}...") try: classifier = pipeline("text-classification", model=MODEL_NAME) print("Model loaded successfully!") except Exception as e: print(f"Error loading model: {e}") classifier = None # ========================================== # PREDICTION LOGIC # ========================================== def predict_single(text): """Classify a single headline.""" if not text or not text.strip(): return create_empty_result() if classifier is None: return create_error_result() start = time.time() result = classifier(text.strip())[0] elapsed = (time.time() - start) * 1000 label = result["label"] score = result["score"] is_clickbait = label == "Clickbait" return create_result_html(text.strip(), is_clickbait, score, elapsed) def predict_batch(texts): """Classify multiple headlines (one per line).""" if not texts or not texts.strip(): return "

Enter headlines, one per line.

" if classifier is None: return create_error_result() lines = [line.strip() for line in texts.strip().split("\n") if line.strip()] if not lines: return "

No valid headlines found.

" start = time.time() results = classifier(lines) elapsed = (time.time() - start) * 1000 html_parts = [] clickbait_count = 0 for text, res in zip(lines, results): is_clickbait = res["label"] == "Clickbait" score = res["score"] if is_clickbait: clickbait_count += 1 color = "#dc2626" if is_clickbait else "#16a34a" bg = "#fef2f2" if is_clickbait else "#f0fdf4" icon = "🚨" if is_clickbait else "✅" label_text = "CLICKBAIT" if is_clickbait else "LEGIT" bar_width = int(score * 100) html_parts.append(f"""
{icon} {text} {label_text} {score:.0%}
""") summary_color = "#dc2626" if clickbait_count > len(lines) / 2 else "#16a34a" summary_bg = "#fef2f2" if clickbait_count > len(lines) / 2 else "#f0fdf4" summary = f"""
Batch Analysis
{clickbait_count} / {len(lines)} Clickbait
Processed in {elapsed:.0f}ms
""" return summary + "\n".join(html_parts) # ========================================== # HTML RESULT BUILDERS # ========================================== def create_result_html(text, is_clickbait, score, elapsed_ms): if is_clickbait: main_color = "#dc2626" gradient = "linear-gradient(135deg, #fee2e2, #fecaca, #fca5a5)" icon = "🚨" label = "CLICKBAIT DETECTED" subtitle = "This headline uses manipulative patterns to attract clicks." text_color = "#991b1b" else: main_color = "#16a34a" gradient = "linear-gradient(135deg, #dcfce7, #bbf7d0, #86efac)" icon = "✅" label = "LEGITIMATE NEWS" subtitle = "This headline appears to be genuine and informative." text_color = "#166534" confidence_pct = int(score * 100) return f"""
{icon}
{label}
{subtitle}
Analyzed Headline
"{text}"
Confidence {confidence_pct}%
Model
RoBERTa
Latency
{elapsed_ms:.0f}ms
Tokens
≤128
""" def create_empty_result(): return """
🔍
Awaiting Input
Enter a headline above and click Analyze
""" def create_error_result(): return """
⚠ī¸
Model Not Available
Please wait while the model loads or try refreshing.
""" # ========================================== # CUSTOM CSS # ========================================== CUSTOM_CSS = """ @import url('https://fonts.googleapis.com/css2?family=Inter:wght@400;500;600;700;800&display=swap'); * { font-family: 'Inter', 'Segoe UI', sans-serif !important; } .gradio-container { max-width: 960px !important; margin: 0 auto !important; background: linear-gradient(180deg, #f8fafc 0%, #f1f5f9 50%, #e2e8f0 100%) !important; } .main-header { text-align: center; padding: 40px 20px 20px; } .main-header h1 { background: linear-gradient(135deg, #6366f1, #8b5cf6, #a855f7); -webkit-background-clip: text; -webkit-text-fill-color: transparent; font-size: 2.5rem !important; font-weight: 800 !important; margin-bottom: 8px !important; letter-spacing: -0.5px; } .main-header p { color: #64748b !important; font-size: 15px !important; } .model-badge { display: inline-block; background: linear-gradient(135deg, #ede9fe, #e0e7ff); border: 1px solid #c7d2fe; color: #4f46e5 !important; padding: 6px 16px; border-radius: 24px; font-size: 13px !important; font-weight: 600; letter-spacing: 0.5px; margin-top: 12px; } footer { display: none !important; } .tab-nav button { color: #64748b !important; font-weight: 600 !important; font-size: 14px !important; } .tab-nav button.selected { color: #6366f1 !important; border-color: #6366f1 !important; } """ # ========================================== # GRADIO UI # ========================================== with gr.Blocks( css=CUSTOM_CSS, title="RoBERTa Clickbait Detector", theme=gr.themes.Soft( primary_hue="indigo", secondary_hue="violet", neutral_hue="slate", ), ) as demo: # Header gr.HTML("""

đŸŽ¯ Clickbait Detector

AI-powered headline analysis built on RoBERTa (125M parameters)

🧠 ENTUM-AI / roberta-clickbait-classifier
""") with gr.Tabs(): # --- Tab 1: Single Analysis --- with gr.Tab("🔍 Single Analysis"): with gr.Row(): with gr.Column(scale=3): single_input = gr.Textbox( label="Headline", placeholder="e.g. You Won't BELIEVE What This Celebrity Did Next!", lines=2, max_lines=3, ) single_btn = gr.Button("⚡ Analyze", variant="primary", size="lg") with gr.Column(scale=4): single_output = gr.HTML(value=create_empty_result()) gr.Examples( examples=[ ["You Won't BELIEVE What This Celebrity Did Next! 😱"], ["Federal Reserve raises interest rates by 0.25 percentage points"], ["10 Shocking Secrets Your Doctor Doesn't Want You to Know"], ["Apple reports Q3 revenue of $81.4 billion, up 2% year over year"], ["This Simple Trick Will Make You a Millionaire Overnight!"], ["The European Central Bank holds interest rates unchanged at 4.5%"], ["SHOCKING: She walked into the room and what happened next changed everything"], ["NASA successfully launches Artemis II mission to lunar orbit"], ], inputs=single_input, label="📋 Try these examples", ) single_btn.click(fn=predict_single, inputs=single_input, outputs=single_output) single_input.submit(fn=predict_single, inputs=single_input, outputs=single_output) # --- Tab 2: Batch Analysis --- with gr.Tab("📊 Batch Analysis"): gr.Markdown("Paste multiple headlines — **one per line** — for batch classification.") with gr.Row(): with gr.Column(scale=2): batch_input = gr.Textbox( label="Headlines (one per line)", placeholder="Headline 1\nHeadline 2\nHeadline 3", lines=8, max_lines=20, ) batch_btn = gr.Button("⚡ Analyze All", variant="primary", size="lg") with gr.Column(scale=3): batch_output = gr.HTML( value="

Results will appear here.

" ) batch_btn.click(fn=predict_batch, inputs=batch_input, outputs=batch_output) # --- Tab 3: About --- with gr.Tab("ℹī¸ About"): gr.HTML("""

About This Model

Architecture RoBERTa-base (125M parameters)
Task Binary Classification (Clickbait / Non-Clickbait)
Language English
Max Input 128 tokens
License Apache 2.0

📊 Training Data

Trained on ~48K samples from three combined & deduplicated English datasets: christinacdl/Clickbait_New, marksverdhei/clickbait_title_classification, and contemmcm/clickbait.

🐍 Python API

from transformers import pipeline

classifier = pipeline("text-classification",
                       model="ENTUM-AI/roberta-clickbait-classifier")

result = classifier("You Won't BELIEVE What Happened!")
# [{'label': 'Clickbait', 'score': 0.99}]

💡 Use Cases

📰
News Aggregators
Filter low-quality clickbait articles
🌐
Social Media
Content moderation & feed quality
🔌
Browser Extensions
Warn users about misleading headlines
📧
Email Filters
Detect clickbait-style subject lines
""") # Launch demo.launch()