Datasets:
metadata
dataset_info:
features:
- name: image
dtype: image
- name: text
dtype: string
splits:
- name: train
num_bytes: 1438457626.5
num_examples: 24508
download_size: 1438550844
dataset_size: 1438457626.5
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
license: apache-2.0
task_categories:
- image-to-text
language:
- sv
tags:
- OCR
- blackletter
- swedish
- newspapers
size_categories:
- 10K<n<100K
Swedish Fraktur
This is a dataset for swedish blackletter from the 19th century. The transcriptions were made by Språkbanken and converted into a text-line dataset by the Swedish National Archives
Dataset Details
Dataset Description
- Curated by: Språkbanken, The Swedish National Archives
- Language(s) (NLP): Swedish
- License: [More Information Needed]
Uses
Direct Use
Train textline-based OCR models for swedish 19th century blackletter
Dataset Structure
{ "image": Image(), "text": str }
Dataset Creation
Source Data
Original dataset from Språkbanken - svenska tidningar 1818-1870
Who are the source data producers?
Dataset Card Authors
Erik Lenas