LiquidAI/structured-extraction
Dataset · https://huggingface.co/datasets/LiquidAI/structured-extraction
Structured Extraction [!Note] 📓 Tutorial: This is the dataset used in SFT with TRL Documents paired with a JSON Schema and the JSON object extracted from them. The documents (emails, receipts, bookings, forms, letters, listings, ...) cover 35 document types in 16 languages. All people, companies, addresses and identifiers are fictional. Splits Split Rows Content train 6,000 30 document types test 500 the same 30 document types, other documents… See the full description on the dataset page: https://huggingface.co/datasets/LiquidAI/structured-extraction.
First seen by Oriel on .
AI summary
The LiquidAI/structured-extraction dataset pairs multilingual documents with JSON Schemas and target JSON extractions for training and evaluating structured extraction models with TRL.
- It covers emails, receipts, bookings, forms, letters, listings, and other documents across 35 document types and 16 languages.
- The train split contains 6,000 rows across 30 document types, while test contains 500 rows covering the same types with different documents.
- The test_unseen split contains 300 rows from five document types absent from training, with six showcase examples.
- Each row includes a document type, language, JSON Schema, document text, and extraction serialized as JSON.
- Extraction conventions standardize dates, times, amounts, currencies, missing values, and selection among multiple candidate values; all entities are fictional.
#structured extraction #json schema #trl #multilingual datasets
Generated by gpt-5.6-sol. Check the original for details.
Metric trends
Last 30 days · Up to 120 recorded samples per source and metric, with reposts kept separate · First and latest retained
Likes
Total likes reported by the source.
- Hugging Face Datasets: 15, latest observation ; +1 over observed period since
Downloads (30d)
Rolling 30-day downloads. A change in this value is not new downloads during the observed period.
- Hugging Face Datasets: 489, latest observation ; +60 over observed period since
Aliases
| Source | External ID | URL |
|---|---|---|
| Hugging Face Datasets | LiquidAI/structured-extraction | https://huggingface.co/datasets/LiquidAI/structured-extraction Read discussion |