Synthetic Invoice PDFs with JSON and Text Annotations

100 fully synthetic, machine-readable invoice PDFs are provided, each paired with JSON and raw text files. The dataset is hosted on Kaggle, but the author, organization, and license details are unknown. The last update date is also unspecified.

Use Cases

• Training OCR models based on the availability of paired PDF and text files.
• Benchmarking document parsing algorithms based on the structured JSON annotations.
• Developing synthetic data generation pipelines for document digitization based on the fully synthetic nature of the dataset.

Strengths

• Contains 100 distinct invoice samples.
• Each invoice is provided in three formats: PDF, JSON, and raw text, enabling multimodal analysis.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.

Provenance

Source

Kaggle

Collection Method

Synthetically generated.