README.md

metadata

language:
  - vi
license:
  - mit
task_categories:
  - document-question-answering
  - image-classification
task_ids:
  - document-question-answering
  - image-captioning
pretty_name: Financial Report Dataset Demo
size_categories:
  - n<1K
tags:
  - financial-reports
  - vietnamese
  - document-processing
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
dataset_info:
  features:
    - name: text
      dtype: string
    - name: image
      dtype: image
    - name: custom_id
      dtype: string
    - name: page_number
      dtype: int64
    - name: report_id
      dtype: string
  splits:
    - name: train
      num_bytes: 46262105
      num_examples: 401
  download_size: 45916378
  dataset_size: 46262105

Dataset Card for Financial Report Dataset Demo

Dataset Description

• Homepage: https://huggingface.co/datasets/kiethuynhanh/financial-report-dataset-demo
• Point of Contact: kiethuynhanh

Dataset Summary

This dataset contains financial reports from Vietnamese VININDEX, including both the text content and corresponding page images. The dataset is designed for document understanding and information extraction tasks.

Languages

The dataset contains text in Vietnamese (vi).

Dataset Structure

The dataset contains 401 examples, each with:

• image: The page image from the PDF document
• text: The extracted text content from the corresponding page
• custom_id: A unique identifier for each example
• page_number: The page number in the original document
• report_id: The unique identifier of the report

Data Fields

• image: A PIL Image object containing the page image
• text: String containing the extracted text content
• custom_id: String identifier in format 'bank/year/quarter/report-page'
• page_number: Integer representing the page number
• report_id: String identifier of the source report

Data Splits

The dataset is provided as a single split containing all examples.

Additional Information

Dataset Creation

The dataset was created by:

1. Extracting text from financial report PDFs
2. Converting PDF pages to images
3. Aligning text content with corresponding page images
4. Validating and cleaning the extracted data

Licensing Information

This dataset is released under MIT License.

Citation Information

Please cite this dataset as:

@misc{financial-report-dataset,
  author       = {Kiet Huynh Anh},
  title        = {Financial Report Dataset Demo},
  year         = {2024},
  publisher    = {Hugging Face},
  howpublished = {https://huggingface.co/datasets/kiethuynhanh/financial-report-dataset-demo}
}