README.md

---
language:
- en
license: apache-2.0
size_categories:
- 10K<n<100K
task_categories:
- token-classification
pretty_name: buster
tags:
- finance
configs:
- config_name: default
  data_files:
  - split: FOLD_1
    path: data/FOLD_1-*
  - split: FOLD_2
    path: data/FOLD_2-*
  - split: FOLD_3
    path: data/FOLD_3-*
  - split: FOLD_4
    path: data/FOLD_4-*
  - split: FOLD_5
    path: data/FOLD_5-*
  - split: SILVER
    path: data/SILVER-*
dataset_info:
  features:
  - name: document_id
    dtype: string
  - name: text
    dtype: string
  - name: tokens
    sequence: string
  - name: labels
    sequence: string
  splits:
  - name: FOLD_1
    num_bytes: 13597946
    num_examples: 753
  - name: FOLD_2
    num_bytes: 13477878
    num_examples: 759
  - name: FOLD_3
    num_bytes: 13602552
    num_examples: 758
  - name: FOLD_4
    num_bytes: 13834760
    num_examples: 755
  - name: FOLD_5
    num_bytes: 13632431
    num_examples: 754
  - name: SILVER
    num_bytes: 111769291
    num_examples: 6196
  download_size: 47212151
  dataset_size: 179914858
---



# Dataset Card for BUSTER
BUSiness Transaction Entity Recognition dataset. 

BUSTER is an Entity Recognition (ER) benchmark for entities related to business transactions. It consists of a gold corpus of 
3779 manually annotated documents on financial transactions that were randomly divided into 5 folds,
plus an additional silver corpus of 6196 automatically annotated documents that were created by the model-optimized RoBERTa system.

### Data Splits Statistics
<table border="1" cellspacing="0" cellpadding="5" style="border-collapse: collapse; width: 100%;">
    <thead>
        <tr>
            <th></th>
            <th></th>
            <th colspan="6" style="text-align:center;">Gold</th>
            <th>Silver</th>
        </tr>
        <tr>
            <th></th>
            <th></th>
            <th>fold 1</th>
            <th>fold 2</th>
            <th>fold 3</th>
            <th>fold 4</th>
            <th>fold 5</th>
            <th>Total</th>
            <th>Total</th>
        </tr>
    </thead>
    <tbody>
        <tr>
            <td></td>
            <td>N. Docs</td>
            <td>753</td>
            <td>759</td>
            <td>758</td>
            <td>755</td>
            <td>754</td>
            <td>3779</td>
            <td>6196</td>
        </tr>
        <tr>
            <td></td>
            <td>N. Tokens</td>
            <td>685K</td>
            <td>680K</td>
            <td>687K</td>
            <td>697K</td>
            <td>688K</td>
            <td>3437K</td>
            <td>5647K</td>
        </tr>
        <tr>
            <td></td>
            <td>N. Annotations</td>
            <td>4119</td>
            <td>4267</td>
            <td>4100</td>
            <td>4103</td>
            <td>4163</td>
            <td>20752</td>
            <td>33272</td>
        </tr>
    </tbody>
</table>



### Pre-print
You can find the pre-print [here](https://arxiv.org/abs/2402.09916).

### Citation Information
If you use BUSTER in your work, please cite us: 

```
@inproceedings{zugarini-etal-2023-buster,
    title = "{BUSTER}: a {``}{BUS}iness Transaction Entity Recognition{''} dataset",
    author = "Zugarini, Andrea  and
      Zamai, Andrew  and
      Ernandes, Marco  and
      Rigutini, Leonardo",
    editor = "Wang, Mingxuan  and
      Zitouni, Imed",
    booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing: Industry Track",
    month = dec,
    year = "2023",
    address = "Singapore",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2023.emnlp-industry.57",
    doi = "10.18653/v1/2023.emnlp-industry.57",
    pages = "605--611",
    abstract = "Albeit Natural Language Processing has seen major breakthroughs in the last few years, transferring such advances into real-world business cases can be challenging. One of the reasons resides in the displacement between popular benchmarks and actual data. Lack of supervision, unbalanced classes, noisy data and long documents often affect real problems in vertical domains such as finance, law and health. To support industry-oriented research, we present BUSTER, a BUSiness Transaction Entity Recognition dataset. The dataset consists of 3779 manually annotated documents on financial transactions. We establish several baselines exploiting both general-purpose and domain-specific language models. The best performing model is also used to automatically annotate 6196 documents, which we release as an additional silver corpus to BUSTER.",
}
```