This dataset is published under the CC-BY 4.0 License.
To cite this dataset:
Chagué, A. (2023). moonshines (Version 2.0.0) [Data set]. https://github.com/alix-tz/moonshines
This dataset is composed of pages of text written in 2023 by a single person, copying texts taken from Guillaume Apollinaire's poems published in Alcools.
The dataset is divided into two parts:
data/which is intended to train transcription models,test/which is intended for test.
The transcription strictly follows what is written on the images, including accentuation or capitalization errors.
The segmentation follows the SegmOnto ontology and mostly relies on MainZone and DefaultLine.
Since the text follows the structure of Alcools, there is almost no ponctuation in this ground truth. Besides, most of the lines start with a capital letter.