|
--- |
|
tags: |
|
- OCR |
|
- HTR |
|
pretty_name: Tschörmen |
|
license: apache-2.0 |
|
task_categories: |
|
- sequence-prediction |
|
language: |
|
- de |
|
--- |
|
|
|
# Tschörmen |
|
A collection of manually handwritten random German text chunks, i.e. |
|
text snippets, words, word snippets. |
|
|
|
The text data originated from [WORTSCHATZ LEIPZIG](https://downloads.wortschatz.uni-leipzig.de/deu_mixed-typical_2021_1M-sentences.txt.gz), the words and word snippets are the |
|
150 most common and 150 least common words and snippets within the base data. |
|
|
|
+++ insert frequency plot +++ |
|
|
|
## Description |
|
This string data can be found under the `labels` column. |
|
|
|
Apart from the labels we provide two different annotations both collected at the same time, representing the same handwriting in two forms |
|
- `input_svg`: svg plot of final annotation (black on white) |
|
- `input_coordinates`: pen stroke coordinates as array of strokes (array of coordinates) |
|
|
|
## Dataset collection |
|
The dataset was collected using the [alphabet](https://easybits-alphabet.netlify.app/) application created by the people of [easybits.tech](https://easybits.tech). |
|
|
|
All samples were submitted by a single annotator within a timespan of a week. |
|
|
