irongateprd/support-ticket-intents

Support Ticket Intents (POC)

Synthetic customer support ticket dataset for training intent routers and sentiment classifiers. For testing and scanner POCs only — all ticket text is fabricated.

Dataset summary

Split	Rows	Purpose
train	1,200	Model fine-tuning
validation	150	Hyperparameter tuning
test	150	Holdout evaluation

Fields

Column	Type	Description
ticket_id	string	Stable identifier (TKT-2026-00001)
subject	string	Ticket subject line (model input)
body	string	Full ticket body (optional second input)
intent	class	Routing label: billing, technical, account, sales, abuse
sentiment	class	CSAT proxy: negative, neutral, positive
priority	string	P1–P4
channel	string	email, chat, phone, portal
created_at	timestamp	ISO-8601 UTC

Label distribution (train)

intent:
  billing     24%
  technical   31%
  account     18%
  sales       20%
  abuse        7%

sentiment:
  negative    22%
  neutral     41%
  positive    37%

Usage

from datasets import load_dataset

ds = load_dataset("matt-ts/support-ticket-intents")
example = ds["train"][0]
print(example["subject"], example["intent"], example["sentiment"])

Provenance

Generated with templated scenarios for Acme Corp POC environments. No real customer data is included.

Models using this dataset

• matt-ts/acme-sentiment-v2
• matt-ts/internal-ticket-router