Datasets:
Formats:
parquet
Languages:
English
Size:
1K - 10K
ArXiv:
Tags:
document-understanding
invoice
ocr
noise-augmentation
confidence-calibration
information-extraction
License:
Cite RealKIE-FCC-Verified and original RealKIE benchmark, clarify dataset lineage
Browse files
README.md
CHANGED
|
@@ -12,6 +12,7 @@ annotations_creators:
|
|
| 12 |
- expert-generated
|
| 13 |
source_datasets:
|
| 14 |
- https://indicodatasolutions.github.io/RealKIE/
|
|
|
|
| 15 |
tags:
|
| 16 |
- document-understanding
|
| 17 |
- invoice
|
|
@@ -293,7 +294,12 @@ This dataset is intended for:
|
|
| 293 |
|
| 294 |
## Source Data
|
| 295 |
|
| 296 |
-
The
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 297 |
|
| 298 |
---
|
| 299 |
|
|
@@ -311,14 +317,27 @@ If you use this dataset, please cite:
|
|
| 311 |
}
|
| 312 |
```
|
| 313 |
|
| 314 |
-
Also cite the
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 315 |
|
| 316 |
```bibtex
|
| 317 |
-
@
|
| 318 |
-
title
|
| 319 |
-
author
|
| 320 |
-
|
| 321 |
-
|
|
|
|
| 322 |
}
|
| 323 |
```
|
| 324 |
|
|
|
|
| 12 |
- expert-generated
|
| 13 |
source_datasets:
|
| 14 |
- https://indicodatasolutions.github.io/RealKIE/
|
| 15 |
+
- https://huggingface.co/datasets/amazon-agi/RealKIE-FCC-Verified
|
| 16 |
tags:
|
| 17 |
- document-understanding
|
| 18 |
- invoice
|
|
|
|
| 294 |
|
| 295 |
## Source Data
|
| 296 |
|
| 297 |
+
The 75 clean FCC invoices used in this dataset are taken from [**RealKIE-FCC-Verified**](https://huggingface.co/datasets/amazon-agi/RealKIE-FCC-Verified), which re-annotated the FCC Invoices subset of the original [**RealKIE**](https://indicodatasolutions.github.io/RealKIE/) benchmark ([Townsend et al., 2024](https://arxiv.org/abs/2403.20101)) to fix two issues in the original annotations:
|
| 298 |
+
|
| 299 |
+
1. **Line Item Grouping** — fields previously treated as independent entries are grouped within each individual line item, aligning annotations with real-world invoice structure.
|
| 300 |
+
2. **Annotation Corrections** — erroneous values in the original annotations were corrected.
|
| 301 |
+
|
| 302 |
+
Starting from these 75 verified documents and their corrected `gt.json` ground truth, this dataset adds noise augmentation using the [Augraphy](https://github.com/sparkfish/augraphy) library with OCR-safe parameter settings, producing up to 18 degraded variants per document for confidence calibration and OCR robustness research.
|
| 303 |
|
| 304 |
---
|
| 305 |
|
|
|
|
| 317 |
}
|
| 318 |
```
|
| 319 |
|
| 320 |
+
Also cite RealKIE-FCC-Verified, which supplied the 75 verified source documents:
|
| 321 |
+
|
| 322 |
+
```bibtex
|
| 323 |
+
@misc{realkiefccverified2026,
|
| 324 |
+
title = {RealKIE-FCC-Verified},
|
| 325 |
+
author = {Amazon AGI},
|
| 326 |
+
year = {2026},
|
| 327 |
+
note = {Dataset available at Hugging Face},
|
| 328 |
+
url = {https://huggingface.co/datasets/amazon-agi/RealKIE-FCC-Verified}
|
| 329 |
+
}
|
| 330 |
+
```
|
| 331 |
+
|
| 332 |
+
And the original RealKIE benchmark:
|
| 333 |
|
| 334 |
```bibtex
|
| 335 |
+
@article{realkie2024,
|
| 336 |
+
title = {RealKIE: Five Novel Datasets for Enterprise Key Information Extraction},
|
| 337 |
+
author = {Townsend, Benjamin and May, Madison and Mackowiak, Katherine and Wells, Christopher},
|
| 338 |
+
journal = {arXiv preprint arXiv:2403.20101},
|
| 339 |
+
year = {2024},
|
| 340 |
+
url = {https://indicodatasolutions.github.io/RealKIE/}
|
| 341 |
}
|
| 342 |
```
|
| 343 |
|