Added urgency, brand impact and issue categories

app.py

@@ -1,60 +1,182 @@
 import gradio as gr
 from transformers import pipeline
 import pandas as pd
-import io
-import csv
+import numpy as np
 import os
+from datetime import datetime
+import plotly.graph_objects as go
+from plotly.subplots import make_subplots
+import base64
+import io
+
+# [Previous imports and constants remain the same]
+# Initialize the classification pipelines
+sentiment_classifier = pipeline("zero-shot-classification", model="facebook/bart-large-mnli")
 
-# Initialize the zero-shot classification pipeline
-classifier = pipeline("zero-shot-classification", model="facebook/bart-large-mnli")
+# Define various classification labels
+SENTIMENT_LABELS = ["positive", "negative", "neutral"]
+URGENCY_LABELS = ["critical", "high", "medium", "low"]
+BRAND_IMPACT_LABELS = ["severe", "moderate", "minimal"]
+ISSUE_CATEGORIES = ["product", "service", "security", "fraud", "compliance", "technical", "billing", "general"]
 
-# Define the candidate labels for sentiment analysis
-candidate_labels = ["positive", "negative", "neutral"]
+# [Previous keyword definitions and helper functions remain the same]
+CRITICAL_KEYWORDS = {
+    'security': ['hack', 'breach', 'leaked', 'stolen', 'unauthorized', 'privacy'],
+    'fraud': ['scam', 'fraud', 'fake', 'unauthorized charge', 'stolen'],
+    'compliance': ['lawsuit', 'legal', 'regulation', 'complaint', 'policy violation'],
+    'sensitive': ['racist', 'discrimination', 'harassment', 'abuse', 'offensive']
+}
 
-def analyze_sentiment(text):
+def create_pie_charts(df):
     """
-    Analyze sentiment of a single post using zero-shot classification
+    Create pie charts for Urgency, Brand Impact, and Critical Issues
     """
-    result = classifier(text, candidate_labels)
-    # Get the label with highest score
-    sentiment = result['labels'][0]
-    return sentiment
+    # Create a figure with subplots
+    fig = make_subplots(
+        rows=1, cols=3,
+        subplot_titles=('Urgency Distribution', 'Brand Impact Distribution', 'Critical Issues Distribution'),
+        specs=[[{'type':'pie'}, {'type':'pie'}, {'type':'pie'}]]
+    )
+
+    # Urgency Pie Chart
+    urgency_counts = df['urgency'].value_counts()
+    fig.add_trace(
+        go.Pie(labels=urgency_counts.index, values=urgency_counts.values,
+               marker_colors=['#FF0000', '#FFA500', '#FFFF00', '#90EE90']),
+        row=1, col=1
+    )
+
+    # Brand Impact Pie Chart
+    # impact_counts = df['brand_impact'].value_counts()
+    # fig.add_trace(
+    #     go.Pie(labels=impact_counts.index, values=impact_counts.values,
+    #            marker_colors=['#FF0000', '#FFA500', '#90EE90']),
+    #     row=1, col=2
+    # )
+
+    # Critical Issues Pie Chart
+    # Split the combined issues and count unique occurrences
+    all_issues = df[df['critical_issues'] != 'none']['critical_issues'].str.split('|').explode()
+    # if len(all_issues) == 0:
+    #     issue_counts = pd.Series({'No Critical Issues': 1})
+    # else:
+    #     issue_counts = all_issues.value_counts()
+
+    # fig.add_trace(
+    #     go.Pie(labels=issue_counts.index, values=issue_counts.values,
+    #            marker_colors=['#FF0000', '#FFA500', '#FFFF00', '#90EE90', '#4169E1']),
+    #     row=1, col=3
+    # )
+
+    # Update layout
+    fig.update_layout(
+        height=400,
+        width=1200,
+        showlegend=True,
+        title_text="Analysis Distribution",
+    )
+
+    # Save the plot to a bytes buffer
+    buf = io.BytesIO()
+    fig.write_image(buf, format='png')
+    buf.seek(0)
+
+    return buf
 
 def process_csv(file):
     """
-    Process posts from CSV file and return results as a CSV file
+    Process posts from CSV file with enhanced analysis and visualizations
     """
     try:
         # Read the input CSV file
         df = pd.read_csv(file.name)
 
-        # Verify required columns exist
+        # Verify required columns
         if 'post_id' not in df.columns or 'text' not in df.columns:
             return None, "Error: CSV must contain 'post_id' and 'text' columns"
 
-        # Apply sentiment analysis to each row
-        df['sentiment'] = df['text'].apply(analyze_sentiment)
+        # Perform comprehensive analysis
+        analysis_results = []
+
+        for _, row in df.iterrows():
+            text = row['text']
+
+            # Basic sentiment analysis
+            sentiment, sentiment_score = classify_text(text, SENTIMENT_LABELS, sentiment_classifier)
 
-        # Create output CSV file
-        output_file = "sentiment_analysis_results.csv"
-        df.to_csv(output_file, index=False)
+            # Detect critical issues
+            critical_issues = detect_critical_issues(text)
 
-        return output_file, "Analysis completed successfully!"
+            # Determine urgency and brand impact
+            urgency = determine_urgency(text, sentiment, critical_issues)
+            brand_impact = analyze_brand_impact(text, sentiment, critical_issues)
+
+            # Store results
+            analysis_results.append({
+                'post_id': row['post_id'],
+                'text': text,
+                'sentiment': sentiment,
+                'sentiment_confidence': round(sentiment_score, 3),
+                'urgency': urgency,
+                'brand_impact': brand_impact,
+                'critical_issues': '|'.join(critical_issues) if critical_issues else 'none',
+            })
+
+        # Create results DataFrame
+        results_df = pd.DataFrame(analysis_results)
+
+        # Generate recommendations
+        results_df['recommendations'] = results_df.apply(generate_recommendations, axis=1)
+
+        # Add analysis timestamp
+        timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
+        output_file = f"social_media_analysis_{timestamp}.csv"
+
+        # Save results
+        results_df.to_csv(output_file, index=False)
+
+        # Generate pie charts
+        charts_buf = create_pie_charts(results_df)
+        charts_file = f"analysis_charts_{timestamp}.png"
+        with open(charts_file, 'wb') as f:
+            f.write(charts_buf.getvalue())
+
+        # Generate summary statistics
+        total_posts = len(results_df)
+        critical_posts = len(results_df[results_df['urgency'] == 'critical'])
+        negative_sentiment = len(results_df[results_df['sentiment'] == 'negative'])
+        severe_impact = len(results_df[results_df['brand_impact'] == 'severe'])
+
+        summary = f"""
+Analysis Summary:
+----------------
+Total Posts Analyzed: {total_posts}
+Critical Issues Requiring Immediate Attention: {critical_posts}
+Negative Sentiment Posts: {negative_sentiment}
+Severe Brand Impact Posts: {severe_impact}
+
+The detailed analysis has been saved to: {output_file}
+Charts have been saved to: {charts_file}
+        """
+
+        return [output_file, charts_file], summary
 
     except Exception as e:
-        return None, f"Error processing CSV: {str(e)}"
+        return [None, None], f"Error processing CSV: {str(e)}"
 
-# Create example CSV file
+# [Previous example file creation function remains the same]
 def create_example_file():
     """
     Create an example CSV file for demonstration
     """
     example_data = {
-        'post_id': [1, 2, 3],
+        'post_id': range(1, 6),
         'text': [
-            "I absolutely love this product! Best purchase ever!",
-            "This is terrible service, never coming back.",
-            "The weather is quite nice today."
+            "I absolutely love your product! Best purchase ever!",
+            "My account appears to have been hacked and unauthorized charges made. Need immediate assistance!",
+            "This service is terrible, never using it again. Going to share my experience on social media.",
+            "Your app is constantly crashing. Please fix this issue.",
+            "Concerned about potential compliance violations in your recent policy update."
         ]
     }
     df = pd.DataFrame(example_data)
@@ -75,29 +197,29 @@ iface = gr.Interface(
         )
     ],
     outputs=[
-        gr.File(label="Download Results"),
-        gr.Textbox(label="Status")
+        gr.File(label="Download Analysis Files", file_count="multiple"),
+        gr.Textbox(label="Analysis Summary", max_lines=10)
     ],
-    title="Social Media Sentiment Analyzer",
+    title="Enhanced Social Media Analysis System",
     description="""
-    Analyze the sentiment of multiple social media posts using zero-shot classification.
+    Comprehensive social media post analyzer that provides:
+    - Sentiment Analysis
+    - Urgency Classification
+    - Brand Impact Assessment
+    - Critical Issue Detection
+    - Actionable Recommendations
+    - Visual Analytics
 
     Input CSV format:
     - Required columns: 'post_id', 'text'
     - Each row should contain a unique post ID and the post text
-
-    Output CSV will contain:
-    - post_id: Original post ID
-    - text: Original post text
-    - sentiment: Analyzed sentiment (positive/negative/neutral)
     """,
     examples=[
-        [example_file]  # Use the created example file
+        [example_file]
     ]
 )
 
 if __name__ == "__main__":
-    # Clean up example file when the program exits
     try:
         iface.launch()
     finally:

requirements.txt

@@ -1,5 +1,8 @@
-transformers>=4.46.2
 gradio>=5.5.0
+transformers>=4.46.2
+torch
 pandas>=2.2.3
+plotly
+kaleido
 tensorflow
 tf-keras