Publish validated dataset, card and quality evidence together

README.md

@@ -0,0 +1,99 @@
+---
+license: cc-by-nc-4.0
+task_categories:
+- text-generation
+language:
+- en
+tags:
+- synthetic
+- synthetic-data
+- workflow-records
+- b2b
+- finance
+- privacy-safe
+
+size_categories:
+- 10K<n<100K
+pretty_name: Corporate Expense Audits
+---
+
+# Corporate Expense Audits (Synthetic B2B Dataset Preview)
+
+This synthetic workflow dataset simulates **Corporate Travel & Expense Auditing Logs** for B2B evaluation, internal testing, and prototyping.
+
+## About this Dataset
+
+The daily pipeline uses AI-assisted research and blueprint review, with deterministic Python rendering for bulk records and local validation before publication. Records are variants of configured workflow scenarios. A larger row count does not establish independent real-world cases or demonstrated model-training performance.
+
+* **Verified record count:** 10,000
+* **Privacy-safe:** Contains synthetic names, phone numbers, and company identifiers.
+* **Structured and checked:** Schema-validated JSONL; inspect the audit for duplicate, content, and similarity-check details.
+* **Suggested workflow fit:** Enterprise ERP builders training expense approval bots.
+* **Category:** Finance
+
+## Schema Definition
+
+Each record contains the following fields:
+
+* `ticket_id` (str): Unique synthetic record identifier
+* `employee_name` (str): Pre-assigned Employee Name
+* `expense_category` (str): Meals, Lodging, Entertainment, or Travel
+* `claim_amount` (str): Value claimed in local currency
+* `policy_violation_flag` (str): Flagged or Approved
+* `audit_justification` (str): Detailed reasoning for approving or rejecting the expense based on policy
+
+---
+
+## Get the Commercial Version
+
+For additional synthetic workflow records, the matching private repo is `HaseebDev/corporate_expense_audits-commercial` and is available under a **commercial license**.
+
+* **Payment:** Lemon Squeezy hosted checkout.
+* **Delivery:** Lemon Squeezy checkout with secure commercial delivery after purchase
+* **Commercial package:** 100,000 records, schema documentation, and quality audit report.
+
+**[Open the commercial access page](https://haseebdev.github.io/synthetic-data-checkout/commercial-access.html?niche=corporate_expense_audits&demo_repo=HaseebDev%2Fcorporate_expense_audits&commercial_repo=HaseebDev%2Fcorporate_expense_audits-commercial&checkout_url=https%3A%2F%2Fsyntheticdatasetlab.lemonsqueezy.com%2Fcheckout%2Fbuy%2F431fde4c-a3ff-4587-825b-7ec496b1e98e%3Fcheckout%255Bcustom%255D%255Bdataset%255D%3Dcorporate_expense_audits%26checkout%255Bcustom%255D%255Bdemo_repo%255D%3DHaseebDev%252Fcorporate_expense_audits%26checkout%255Bcustom%255D%255Bcommercial_repo%255D%3DHaseebDev%252Fcorporate_expense_audits-commercial%26utm_source%3Dhuggingface%26utm_medium%3Ddataset_card%26utm_campaign%3Dcorporate_expense_audits&commercial_records=100000)**
+
+## Quality Audit
+
+# Dataset Quality Report
+
+## Executive Summary
+
+| Metric | Value |
+| :--- | :--- |
+| Status | **PASS** |
+| Quality Score | **94/100** |
+| Records Audited | 10000 |
+| Niche | Corporate Travel & Expense Auditing Logs |
+| Category | Finance |
+
+## Quality Gates
+
+| Dimension | Result |
+| :--- | :--- |
+| Schema Conformance | 100.0% (10000/10000) |
+| Duplicate Rows | 0 |
+| Duplicate Identity Values | 0 |
+| Non-Synthetic Phone Risk | 0 |
+| Email Address Count | 0 |
+| Max Fuzzy Similarity | 0.5 |
+| Average Words Per Descriptive Field | 94.59 |
+| Vocabulary Diversity | 0.759 |
+| Repeated Phrase Count | 78 |
+| Domain Keyword Coverage | 1.0 |
+
+## Buyer Assurance
+
+This dataset was checked with deterministic local tooling before publication. The audit verifies schema consistency, uniqueness, text depth, duplicate risk, repeated wording, domain terminology coverage, and obvious PII leakage patterns.
+
+All rows receive schema and content checks. Cross-record similarity uses a bounded sample of pairs to keep generation fast; it does not establish that every possible pair is dissimilar. The score measures these automated checks, not downstream model performance or independent medical, legal, or regulatory validation.
+
+## Findings
+
+- **MINOR**: 78 high-frequency non-domain four-word phrase pattern(s) detected.
+
+
+## Support
+
+**Add me on Discord: `xomohappy` for access support, delivery questions, or product questions about this premade commercial dataset.**

quality-audit.json

@@ -0,0 +1,117 @@
+{
+  "niche_id": "corporate_expense_audits",
+  "niche_name": "Corporate Travel & Expense Auditing Logs",
+  "category": "Finance",
+  "record_count": 10000,
+  "expected_fields": [
+    "ticket_id",
+    "employee_name",
+    "expense_category",
+    "claim_amount",
+    "policy_violation_flag",
+    "audit_justification"
+  ],
+  "descriptive_fields": [
+    "audit_justification"
+  ],
+  "quality_score": 94,
+  "passed": true,
+  "schema": {
+    "valid_records": 10000,
+    "conformance_pct": 100.0,
+    "missing_fields": {},
+    "empty_fields": {},
+    "extra_fields": {}
+  },
+  "uniqueness": {
+    "duplicate_rows": 0,
+    "duplicate_identity_values": 0,
+    "duplicate_fields": {}
+  },
+  "text_depth": {
+    "field_count": 1,
+    "avg_chars": 691.79,
+    "min_chars": 565,
+    "avg_words": 94.59,
+    "min_words": 76,
+    "vocabulary_diversity": 0.759,
+    "total_tokens": 945943,
+    "unique_tokens": 504
+  },
+  "similarity": {
+    "max_similarity": 0.5,
+    "avg_similarity": 0.154,
+    "near_duplicate_pairs": 0
+  },
+  "repetition": {
+    "repeated_ngram_count": 78,
+    "repeated_ngram_rate": 0.057,
+    "top_repeated_phrases": [
+      "the itemized meal receipt",
+      "itemized meal receipt and",
+      "meal receipt and daily",
+      "receipt and daily allowance",
+      "reconcile the excess against"
+    ],
+    "frequency_threshold": 1000,
+    "raw_repeated_ngram_count": 9179,
+    "raw_repeated_ngram_rate": 0.959
+  },
+  "pii": {
+    "email_count": 0,
+    "non_synthetic_email_count": 0,
+    "phone_count": 0,
+    "non_synthetic_phone_count": 0
+  },
+  "domain": {
+    "keyword_set_size": 8,
+    "matched_keywords": [
+      "allowance",
+      "approver",
+      "booking",
+      "claim",
+      "expense",
+      "policy",
+      "receipt",
+      "reimbursement"
+    ],
+    "keyword_coverage": 1.0
+  },
+  "balance": {
+    "fields": {
+      "expense_category": {
+        "unique_values": 4,
+        "top_share": 0.334,
+        "values": {
+          "Meals": 3342,
+          "Travel": 3328,
+          "Lodging": 1666,
+          "Entertainment": 1664
+        }
+      },
+      "policy_violation_flag": {
+        "unique_values": 1,
+        "top_share": 1.0,
+        "values": {
+          "Flagged": 10000
+        }
+      }
+    }
+  },
+  "findings": [
+    {
+      "severity": "minor",
+      "message": "78 high-frequency non-domain four-word phrase pattern(s) detected."
+    }
+  ],
+  "content": {
+    "rejected_record_count": 0,
+    "rejections": []
+  },
+  "scenario_coverage": null,
+  "verification": {
+    "all_rows_checked": 10000,
+    "similarity_is_bounded": true,
+    "max_similarity_pairs": 20000
+  }
+}

quality-audit.md

@@ -0,0 +1,37 @@
+# Dataset Quality Report
+
+## Executive Summary
+
+| Metric | Value |
+| :--- | :--- |
+| Status | **PASS** |
+| Quality Score | **94/100** |
+| Records Audited | 10000 |
+| Niche | Corporate Travel & Expense Auditing Logs |
+| Category | Finance |
+
+## Quality Gates
+
+| Dimension | Result |
+| :--- | :--- |
+| Schema Conformance | 100.0% (10000/10000) |
+| Duplicate Rows | 0 |
+| Duplicate Identity Values | 0 |
+| Non-Synthetic Phone Risk | 0 |
+| Email Address Count | 0 |
+| Max Fuzzy Similarity | 0.5 |
+| Average Words Per Descriptive Field | 94.59 |
+| Vocabulary Diversity | 0.759 |
+| Repeated Phrase Count | 78 |
+| Domain Keyword Coverage | 1.0 |
+
+## Buyer Assurance
+
+This dataset was checked with deterministic local tooling before publication. The audit verifies schema consistency, uniqueness, text depth, duplicate risk, repeated wording, domain terminology coverage, and obvious PII leakage patterns.
+
+All rows receive schema and content checks. Cross-record similarity uses a bounded sample of pairs to keep generation fast; it does not establish that every possible pair is dissimilar. The score measures these automated checks, not downstream model performance or independent medical, legal, or regulatory validation.
+
+## Findings
+
+- **MINOR**: 78 high-frequency non-domain four-word phrase pattern(s) detected.
+