metadata
dataset_info:
features:
- name: image
dtype: image
- name: gt_json
dtype: string
splits:
- name: train
num_bytes: 112840658
num_examples: 109
- name: val
num_bytes: 10941585
num_examples: 10
- name: test
num_bytes: 27477418
num_examples: 26
download_size: 150586805
dataset_size: 151259661
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
- split: val
path: data/val-*
- split: test
path: data/test-*
license: apache-2.0
task_categories:
- text-generation
language:
- pl
tags:
- legal
pretty_name: Polish Insurance Terms Structuring
size_categories:
- n<1K
A dataset for the task of document structuring (parsing) Polish legal documents with nested lists.
The dataset contains 2 columns:
image: pdf pages converted to 1080x1440 images,gt_json: a json containing a list of detected objects undergt_parsekey.
The detected objects in the gt_json column are specified as follows:
| Object Type | Keys | Description |
|---|---|---|
| Heading | content | Represents a heading-like loose block of text. |
| List | title, items | A collection of items, which can be an element, an overflowing element, or another list. |
| Element | content | A basic unit of content within a list (along with the list prefix, e.g. "1.") or a loose block of text. |
| Overflowing Element | content | Content that continues from the main list or element, but isn't part of the standard list flow. |
| Element Continuation | content | Content that continues from an element on the previous page. |
| List Continuation | items | List continuation from the previous page; items can be an element, an overflowing element, or another list. |
All images were obtained from Polish insurance companies official websites. All sources are provided below: