How to use ChatGPT at Scale to Solve Enterprise Problems

				
					import json
import pandas as pd
import os
from collections import Counter
from datetime import datetime

# Path to the folder containing JSON files
apple_folder = "apple_news"
all_data = []

if os.path.exists(apple_folder):
    json_files = [f for f in os.listdir(apple_folder) if f.endswith('.json')]
    
    for file in json_files:

        file_path = os.path.join(apple_folder, file)
        try:
            with open(file_path, 'r') as f:
                data = json.load(f)
                company = data.get("company", "Unknown")
                retrieved_at = data.get("retrieved_at", None)
                for article in data.get("articles", []):
                    article["company"] = company
                    article["retrieved_at"] = retrieved_at
                    all_data.append(article)       
        except Exception as e:
            print(f"Error reading {file}: {e}")
else:
    print(f"Folder '{apple_folder}' does not exist!")

df = pd.DataFrame(all_data)
df['parsed_date'] = pd.to_datetime(df['date'])

# 1. Sentiment Count
sentiment_counts = Counter(df['sentiment'])

# 2. Most Recent Article Per Sentiment
latest_articles_by_sentiment = {}
for sentiment, group in df.groupby('sentiment'):
    latest = group.sort_values('parsed_date', ascending=False).iloc[0]
    latest_articles_by_sentiment[sentiment] = latest.to_dict()

# 3. Top Sources
source_counts = Counter(df['source'])

# 4. Sentiment Over Time (basic trend detection)
timeline_df = df.sort_values('parsed_date')
timeline = list(zip(timeline_df['date'], timeline_df['sentiment']))

# 5. Summary Insight Generator
def generate_insight():
    total = len(df)
    companies = df['company'].unique()
    company_text = ', '.join(companies) if len(companies) <= 3 else f"{companies[0]} and {len(companies)-1} others"
    
    pos = sentiment_counts.get("positive", 0)
    neu = sentiment_counts.get("neutral", 0)
    neg = sentiment_counts.get("negative", 0)
    mixed = sentiment_counts.get("interested", 0)

    insight = f"🔍 Sentiment Report for {company_text} ({total} articles)\n"
    insight += f"- Positive: {pos} ({(pos/total)*100:.1f}%)\n"
    insight += f"- Neutral: {neu} ({(neu/total)*100:.1f}%)\n"
    insight += f"- Negative: {neg} ({(neg/total)*100:.1f}%)\n" if neg > 0 else ""
    insight += f"- Interested (mixed): {mixed} ({(mixed/total)*100:.1f}%)\n" if mixed > 0 else ""

    insight += f"\n📰 Top Sources:\n"
    for source, count in source_counts.most_common(3):
        insight += f"- {source}: {count} articles\n"

    insight += f"\n📅 Recent Coverage by Sentiment:\n"
    for sentiment, article in latest_articles_by_sentiment.items():
        insight += f"- {sentiment.capitalize()}: \"{article['title']}\" ({article['date']})\n"

    # Optional: detect tone shift
    if len(timeline) >= 2:
        first_sentiment = timeline[0][1]
        last_sentiment = timeline[-1][1]
        if first_sentiment != last_sentiment:
            insight += f"\n⚠️ Tone shift detected from '{first_sentiment}' to '{last_sentiment}' over time.\n"

    return insight

# Output result
report = generate_insight()
print(report)