Synthetic Customer Support Tickets with Canary Tokens for Privacy Testing

Yashaswee Narayan's dataset contains 500 synthetically generated customer support tickets. The records simulate realistic interactions and include uniquely identifiable canary tokens. It was last updated on June 3, 2026.

Use Cases

• Evaluating text classification models based on the naturalistic variations in syntax, tone, and issue context.
• Testing for information extraction from unstructured customer support text.
• Measuring model memorization and membership inference based on the presence of injected canary tokens.
• Assessing critical privacy leakage ratios in Large Language Models (LLMs).

Strengths

• Contains 500 records, providing a defined scale for evaluation.
• Includes uniquely identifiable canary tokens (e.g., CANARY-BILL-9921-X) for traceable privacy testing.
• Released under the permissive CC-BY-4.0 license.

Limitations

• Row count is unknown, which may limit suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.
• Data is synthetic, which may not capture all nuances of real-world customer interactions.

Provenance

Source

Yashaswee Narayan via figshare.

Collection Method

Synthetically generated.

Time Range

null

Freshness

Last updated 2026-06-03 13:51:31; freshness should be verified.

Geography

null