Upload scripts/build_synthetic_customer_support_dataset.py with huggingface_hub

scripts/build_synthetic_customer_support_dataset.py

@@ -0,0 +1,196 @@
+import csv
+import json
+import random
+from datetime import date, timedelta
+from pathlib import Path
+
+random.seed(155)
+
+OUT_DIR = Path("data/synthetic_customer_support")
+OUT_DIR.mkdir(parents=True, exist_ok=True)
+
+START_DATE = date(2025, 1, 1)
+
+PRODUCTS = [
+    "AI Assistant SaaS",
+    "ERP Cloud Suite",
+    "Private AI Desktop",
+    "CRM Automation",
+    "Analytics Dashboard",
+    "eCommerce Growth Pack",
+    "HR Recruiter AI",
+    "Document Intelligence API"
+]
+
+CATEGORIES = [
+    "Billing",
+    "Login Issue",
+    "Feature Request",
+    "Bug Report",
+    "Account Setup",
+    "Integration",
+    "Performance",
+    "Data Import",
+    "Subscription",
+    "General Question"
+]
+
+PRIORITIES = ["Low", "Medium", "High", "Urgent"]
+STATUSES = ["Open", "Pending", "Resolved", "Closed"]
+CHANNELS = ["Email", "Chat", "Phone", "Web Form", "In-App"]
+SENTIMENTS = ["Positive", "Neutral", "Negative", "Frustrated"]
+PLANS = ["Free", "Starter", "Professional", "Business", "Enterprise"]
+
+AGENTS = [
+    "Support Agent A",
+    "Support Agent B",
+    "Support Agent C",
+    "Support Agent D",
+    "Support Agent E"
+]
+
+KB_TOPICS = [
+    "How to reset password",
+    "How to update billing method",
+    "How to import CSV data",
+    "How to connect API key",
+    "How to export reports",
+    "How to invite team members",
+    "How to troubleshoot slow dashboard",
+    "How to upgrade subscription",
+    "How to configure integrations",
+    "How to use AI assistant"
+]
+
+def random_date():
+    return START_DATE + timedelta(days=random.randint(0, 364))
+
+def write_csv(path, rows):
+    with path.open("w", encoding="utf-8", newline="") as f:
+        writer = csv.DictWriter(f, fieldnames=list(rows[0].keys()))
+        writer.writeheader()
+        writer.writerows(rows)
+
+def write_jsonl(path, rows):
+    with path.open("w", encoding="utf-8") as f:
+        for row in rows:
+            f.write(json.dumps(row, ensure_ascii=False) + "\n")
+
+kb_rows = []
+for i, topic in enumerate(KB_TOPICS, start=1):
+    kb_rows.append({
+        "article_id": f"KB-{i:05d}",
+        "title": topic,
+        "category": random.choice(CATEGORIES),
+        "product": random.choice(PRODUCTS),
+        "summary": f"Synthetic help article explaining: {topic}.",
+        "recommended_action": random.choice([
+            "Follow step-by-step guide",
+            "Check account settings",
+            "Contact support if issue continues",
+            "Review integration configuration",
+            "Retry after clearing cache"
+        ]),
+        "is_synthetic": True,
+        "source": "Synthetic Realigns customer support dataset",
+        "license": "Synthetic data generated for public developer use"
+    })
+
+tickets = []
+interactions = []
+
+for i in range(1, 2001):
+    ticket_id = f"TICKET-{i:07d}"
+    created_date = random_date()
+    priority = random.choice(PRIORITIES)
+    status = random.choice(STATUSES)
+    category = random.choice(CATEGORIES)
+
+    first_response_hours = {
+        "Low": random.uniform(6, 48),
+        "Medium": random.uniform(2, 24),
+        "High": random.uniform(0.5, 8),
+        "Urgent": random.uniform(0.1, 2),
+    }[priority]
+
+    resolution_hours = first_response_hours + random.uniform(2, 120)
+    if status in ["Open", "Pending"]:
+        resolution_hours = None
+
+    tickets.append({
+        "ticket_id": ticket_id,
+        "created_date": created_date.isoformat(),
+        "customer_plan": random.choice(PLANS),
+        "product": random.choice(PRODUCTS),
+        "category": category,
+        "priority": priority,
+        "status": status,
+        "channel": random.choice(CHANNELS),
+        "assigned_agent": random.choice(AGENTS),
+        "sentiment": random.choice(SENTIMENTS),
+        "first_response_hours": round(first_response_hours, 2),
+        "resolution_hours": round(resolution_hours, 2) if resolution_hours is not None else None,
+        "sla_breached": bool(resolution_hours and resolution_hours > random.choice([24, 48, 72])),
+        "related_kb_article_id": random.choice(kb_rows)["article_id"],
+        "is_synthetic": True,
+        "privacy_note": "No real customers, emails, phone numbers, account IDs, or private support messages are included.",
+        "source": "Synthetic Realigns customer support dataset",
+        "license": "Synthetic data generated for public developer use"
+    })
+
+    for j in range(random.randint(1, 6)):
+        interaction_date = created_date + timedelta(days=random.randint(0, 20))
+        interactions.append({
+            "interaction_id": f"INT-{len(interactions) + 1:08d}",
+            "ticket_id": ticket_id,
+            "interaction_date": interaction_date.isoformat(),
+            "sender_type": random.choice(["Customer", "Support Agent", "System"]),
+            "channel": random.choice(CHANNELS),
+            "message_type": random.choice(["Question", "Reply", "Status Update", "Troubleshooting Step", "Resolution Note"]),
+            "message_summary": random.choice([
+                "Customer reported an issue and requested help.",
+                "Agent provided troubleshooting instructions.",
+                "System updated ticket status.",
+                "Customer confirmed issue still exists.",
+                "Agent shared knowledge base guidance.",
+                "Ticket was resolved after follow-up."
+            ]),
+            "is_synthetic": True,
+            "source": "Synthetic Realigns customer support dataset",
+            "license": "Synthetic data generated for public developer use"
+        })
+
+write_csv(OUT_DIR / "synthetic_support_tickets.csv", tickets)
+write_jsonl(OUT_DIR / "synthetic_support_tickets.jsonl", tickets)
+
+write_csv(OUT_DIR / "synthetic_support_interactions.csv", interactions)
+write_jsonl(OUT_DIR / "synthetic_support_interactions.jsonl", interactions)
+
+write_csv(OUT_DIR / "synthetic_support_knowledge_base.csv", kb_rows)
+write_jsonl(OUT_DIR / "synthetic_support_knowledge_base.jsonl", kb_rows)
+
+metadata = {
+    "dataset": "Synthetic Customer Support Tickets Dataset",
+    "prepared_by": "Realigns Inc.",
+    "records": {
+        "support_tickets": len(tickets),
+        "support_interactions": len(interactions),
+        "knowledge_base_articles": len(kb_rows)
+    },
+    "year": 2025,
+    "privacy": "No real customers, emails, phone numbers, account IDs, support messages, or private customer service records are included.",
+    "intended_use": [
+        "AI customer support assistant testing",
+        "helpdesk dashboard development",
+        "ticket classification",
+        "SLA analytics",
+        "sentiment and priority analysis",
+        "RAG support workflows"
+    ],
+    "license": "Synthetic data generated for public developer use"
+}
+
+(OUT_DIR / "metadata.json").write_text(json.dumps(metadata, indent=2), encoding="utf-8")
+
+print("Done. Synthetic Customer Support dataset created.")
+print(json.dumps(metadata, indent=2))