app.py

import gradio as gr
from transformers import pipeline
import json

# Initialize NLP pipeline
ner_pipeline = pipeline("ner", model="dbmdz/bert-large-cased-finetuned-conll03-english")

def analyze_event(text):
    try:
        # Process text with NER pipeline
        ner_results = ner_pipeline(text)
        
        # Group entities
        entities = {
            "people": [],
            "organizations": [],
            "locations": [],
            "hashtags": [word for word in text.split() if word.startswith('#')]
        }
        
        for item in ner_results:
            if item["entity"].endswith("PER"):
                entities["people"].append(item["word"])
            elif item["entity"].endswith("ORG"):
                entities["organizations"].append(item["word"])
            elif item["entity"].endswith("LOC"):
                entities["locations"].append(item["word"])
        
        # Calculate confidence
        confidence = min(1.0, (
            0.2 * bool(entities["people"]) +
            0.2 * bool(entities["organizations"]) +
            0.3 * bool(entities["locations"]) +
            0.3 * bool(entities["hashtags"])
        ))
        
        return {
            "text": text,
            "entities": entities,
            "confidence": confidence,
            "verification_needed": confidence < 0.6
        }
    except Exception as e:
        return {"error": str(e)}

# Create Gradio interface with custom CSS and HTML
css = """
.container { max-width: 800px; margin: auto; padding: 20px; }
.results { padding: 20px; border: 1px solid #ddd; border-radius: 8px; margin-top: 20px; }
.confidence-high { color: #22c55e; font-weight: bold; }
.confidence-low { color: #f97316; font-weight: bold; }
.entity-section { margin: 15px 0; }
.alert-warning { background: #fff3cd; padding: 10px; border-radius: 5px; margin: 10px 0; }
.alert-success { background: #d1fae5; padding: 10px; border-radius: 5px; margin: 10px 0; }
"""

def format_results(analysis_result):
    if "error" in analysis_result:
        return f"<div style='color: red'>Error: {analysis_result['error']}</div>"
    
    confidence_class = "confidence-high" if analysis_result["confidence"] >= 0.6 else "confidence-low"
    
    html = f"""
    <div class="results">
        <div style="display: flex; justify-content: space-between; align-items: center; margin-bottom: 20px;">
            <h3 style="margin: 0;">Analysis Results</h3>
            <div>
                Confidence Score: <span class="{confidence_class}">{int(analysis_result['confidence'] * 100)}%</span>
            </div>
        </div>
        
        {f'''
        <div class="alert-warning">
            ⚠️ <strong>Verification Required:</strong> Low confidence score detected. Please verify the extracted information.
        </div>
        ''' if analysis_result["verification_needed"] else ''}
        
        <div class="entity-section">
            <h4>👤 People Detected</h4>
            <ul>{''.join(f'<li>{person}</li>' for person in analysis_result['entities']['people']) or '<li>None detected</li>'}</ul>
        </div>
        
        <div class="entity-section">
            <h4>🏢 Organizations</h4>
            <ul>{''.join(f'<li>{org}</li>' for org in analysis_result['entities']['organizations']) or '<li>None detected</li>'}</ul>
        </div>
        
        <div class="entity-section">
            <h4>📍 Locations</h4>
            <ul>{''.join(f'<li>{loc}</li>' for loc in analysis_result['entities']['locations']) or '<li>None detected</li>'}</ul>
        </div>
        
        <div class="entity-section">
            <h4># Hashtags</h4>
            <ul>{''.join(f'<li>{tag}</li>' for tag in analysis_result['entities']['hashtags']) or '<li>None detected</li>'}</ul>
        </div>
        
        {f'''
        <div class="alert-success">
            ✅ <strong>Event Validated:</strong> The extracted information meets confidence thresholds.
        </div>
        ''' if not analysis_result["verification_needed"] else ''}
    </div>
    """
    return html

demo = gr.Interface(
    fn=lambda text: format_results(analyze_event(text)),
    inputs=[
        gr.Textbox(
            label="Event Text",
            placeholder="Enter text to analyze (e.g., 'John from Tech Corp. is attending the meeting in Washington, DC #tech')",
            lines=3
        )
    ],
    outputs=gr.HTML(),
    title="DoD Event Analysis System",
    description="Analyze text to extract entities, assess confidence, and identify key event information.",
    css=css,
    theme=gr.themes.Soft(),
    examples=[
        ["John from Tech Corp. is attending the meeting in Washington, DC tomorrow #tech"],
        ["Sarah Johnson and Mike Smith from Defense Systems Inc. are conducting training in Norfolk, VA #defense #training"],
        ["Team meeting at headquarters with @commander_smith #briefing"]
    ]
)

if __name__ == "__main__":
    demo.launch()