README.md

---
license: mit
size_categories:
- n<1K
---
# Dataset Card for Suzhou Numeral Dataset 🌟
![Sample Image](assets/KM_48690-0002-u.jpg)  
This dataset contains annotated images of Suzhou numerals (0-9) extracted from historical archival records of the Hechang Firm in Nagasaki 📜, documenting trade activities between China, Japan, and Southeast Asia from 1880 to 1930. It’s designed to support research in historical document analysis, optical character recognition (OCR) 🔍, and the study of traditional numeral systems 📊.

## Dataset Details 🎉

### Dataset Description

<!-- Provide a longer summary of what this dataset is. -->  
This dataset comprises **773 annotated instances** of Suzhou numerals derived from the Hechang Firm in Nagasaki archive, a collection of accounting ledgers, trade contracts, and commercial correspondence 💼. The dataset includes high-resolution .png images of Suzhou numerals alongside their corresponding labels in CSV files, capturing natural variations in notation style, stroke thickness, and alignment ✍️. It is split into **training (541 samples)**, **testing (116 samples)**, and **validation (116 samples)** sets with a 7:1.5:1.5 ratio.  
![Sample Image](assets/flow_chart_v8.jpg)  

- **Curated by:** [More Information Needed]  
- **Funded by [optional]:** [More Information Needed]  
- **Shared by [optional]:** [More Information Needed]  
- **Language(s) (NLP):** [More Information Needed]  
- **License:** [More Information Needed]  

### Dataset Sources [optional] 🔗

<!-- Provide the basic links for the dataset. -->  
- **Repository:** [More Information Needed]  
- **Paper [optional]:** [More Information Needed]  
- **Demo [optional]:** [More Information Needed]  

## Uses 🚀

<!-- Address questions around how the dataset is intended to be used. -->  

### Direct Use ✅  
This dataset is suitable for:  
- Training OCR models to recognize Suzhou numerals 🤖.  
- Studying historical variations in numeral notation and document formatting 📚.  
- Benchmarking image classification algorithms on real-world handwritten data 🧪.  
[More Information Needed]  

### Out-of-Scope Use 🚫  
The dataset is **not intended** for:  
- General-purpose digit recognition beyond Suzhou numerals.  
- Applications requiring modern numeral systems or non-historical contexts.  
[More Information Needed]  

## Dataset Structure 🗂️

<!-- This section provides a description of the dataset fields, and additional information about the dataset structure such as criteria used to create the splits, relationships between data points, etc. -->  
```
The dataset is organized as follows:
├── train.csv          # Training set metadata (541 samples)
├── test.csv           # Testing set metadata (116 samples)
├── val.csv            # Validation set metadata (116 samples)
├── images/            # Folder containing 773 individual .png images
│   ├── 12.png         # Example image file
│   ├── 13.png
│   └── ...
├── images.zip         # Zipped archive of the images/ folder
```

- train.csv, test.csv, val.csv:
  - Columns:
    - image_name: Filename of the image (e.g., 12.png).
    - label: Suzhou numeral label as a string (e.g., 22455).
  - These files map images to their annotated labels.
- images/: Contains all .png images of cropped Suzhou numerals.
![Sample Image](images/14.png)
- images.zip: A compressed archive of the images/ folder.

The splits were created with a 7:1.5:1.5 ratio to ensure a balanced distribution for training, testing, and validation.
[More Information Needed]

## Dataset Creation ✨

### Curation Rationale  
<!-- Motivation for the creation of this dataset. -->  
This dataset was created to **preserve and analyze** the use of Suzhou numerals in historical trade records, enabling research into traditional numeral systems and supporting the development of OCR tools for historical document digitization 🛠️.  
[More Information Needed]  

### Source Data  

#### Data Collection and Processing  
<!-- This section describes the data collection and processing process such as data selection criteria, filtering and normalization methods, tools and libraries used, etc. -->  
The dataset originates from the Hechang Firm in Nagasaki archive, spanning **1880–1930**. The process involved:  
1. Scanning documents into high-resolution PDFs 📑.  
2. Manually identifying and annotating Suzhou numerals (0-9) within transaction records, cost lists, and handwritten text ✍️.  
3. Cropping annotated sections into individual .png images, each paired with a label 🖼️.  
4. Cross-verifying ambiguous cases (e.g., faded or overlapping strokes) by multiple annotators for consistency ✅.  
[More Information Needed]  

#### Who are the source data producers?  
The source data was produced by **Chinese merchants and traders** associated with the Hechang Firm in Nagasaki, who used Suzhou numerals in their accounting and correspondence 💰. No specific demographic details are available.  
[More Information Needed]  

### Annotations [optional]  
<!-- If the dataset contains annotations which are not part of the initial data collection, use this section to describe them. -->  

#### Annotation process  
Human experts manually annotated Suzhou numerals in the digitized documents. The process included:  
- Identifying numeral instances in PDFs 🔍.  
- Assigning labels based on the depicted numeral ✍️.  
- Cropping images and validating annotations, with multiple annotators resolving ambiguities ✅.  
[More Information Needed]  

#### Who are the annotators?  
The annotators were **human experts** familiar with historical Chinese numerals and document analysis 🧑‍🏫.  
[More Information Needed]  

#### Personal and Sensitive Information  
The dataset contains **no personal or sensitive information**, as it focuses solely on numeral images extracted from historical trade records. No identifiable data (e.g., names, addresses) is included 🚫.  
[More Information Needed]  

## Bias, Risks, and Limitations ⚠️

<!-- This section is meant to convey both technical and sociotechnical limitations. -->  
- **Bias**: The dataset reflects Suzhou numeral usage specific to the Hechang Firm’s records (1880–1930), which may not generalize to other regions or time periods 🌍.  
- **Risks**: Misinterpretation of faded or ambiguous numerals could affect model performance 😕.  
- **Limitations**: Limited sample size (773 instances) and focus on a niche numeral system may restrict broader applicability 📉.  
[More Information Needed]  

### Recommendations 💡  
Users should be aware of the **historical and regional specificity** of the dataset. For robust OCR applications, supplementing with additional Suzhou numeral data from other sources is recommended 🌟.  
Users should be made aware of the risks, biases, and limitations of the dataset. More information needed for further recommendations.  

## Citation [optional] 📖

<!-- If there is a paper or blog post introducing the dataset, the APA and Bibtex information for that should go in this section. -->  

**BibTeX:**  
```bibtex
@inproceedings{wu-etal-2025-bringing,
    title = "Bringing Suzhou Numerals into the Digital Age: A Dataset and Recognition Study on {A}ncient {C}hinese Trade Records",
    author = "Wu, Ting-Lin  and
      Chen, Zih-Ching  and
      Chen, Chen-Yuan  and
      Chen, Pi-Jhong  and
      Wang, Li-Chiao",
    editor = "Anderson, Adam  and
      Gordin, Shai  and
      Li, Bin  and
      Liu, Yudong  and
      Passarotti, Marco C.  and
      Sprugnoli, Rachele",
    booktitle = "Proceedings of the Second Workshop on Ancient Language Processing",
    month = may,
    year = "2025",
    address = "The Albuquerque Convention Center, Laguna",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2025.alp-1.13/",
    pages = "105--111",
    ISBN = "979-8-89176-235-0",
    abstract = "Suzhou numerals, a specialized numerical no-tation system historically used in Chinese com-merce and accounting, played a pivotal role in financial transactions from the Song Dynasty to the early 20th century. Despite their his-torical significance, they remain largely absent from modern OCR benchmarks, limiting com-putational access to archival trade documents. This paper presents a curated dataset of 773 expert-annotated Suzhou numeral samples ex-tracted from late Qing-era trade ledgers. We provide a statistical analysis of character distri-butions, offering insights into their real-world usage in historical bookkeeping. Additionally, we evaluate baseline performance with hand-written text recognition (HTR) model, high-lighting the challenges of recognizing low-resource brush-written numerals. By introduc-ing this dataset and initial benchmark results, we aim to facilitate research in historical doc-umentation in ancient Chinese characters, ad-vancing the digitization of early Chinese finan-cial records. The dataset is publicly available at our huggingface hub, and our codebase can be accessed at our github repository."
}
```

**APA:**
[More Information Needed]. (2025). Suzhou Numeral Dataset. Hugging Face. [More Information Needed]
[More Information Needed]

## Glossary [optional]

<!-- If relevant, include terms and calculations in this section that can help readers understand the dataset or dataset card. -->

[More Information Needed]

## More Information [optional]

[More Information Needed]

## Dataset Card Authors [optional]

[More Information Needed]

## Dataset Card Contact

[More Information Needed]