---
language:
- en
pretty_name: Wiki12k
size_categories:
- 10K<n<100K
license: cc-by-sa-4.0
---

Subset of 12.500 Wikipedia files extracted from the Wiki727k Text Segmentation dataset for my master thesis.
Distribution of Wikipedia Content is managed under the Creative Commons Attribution-ShareAlike 4.0 International (CC BY-SA 4.0) License.
Text Segmentation is the task of splitting a text into semantic coherent paragraphs. 
Each file has a number of paragraphs, divided by '========,\<number\>,\<title\>'
I suggest to use this dataset via cloning with `git clone https://huggingface.co/datasets/mathieuschanz/Wiki12kTextSegmentation`, 
because I encountered problems by downloading the dataset with the Huggingface library. 
I also made slight adjustments for my usecases to the Koomri's loading scripts. Coming perhaps soon. 

For the whole dataset, the Choi dataset, or the loading script, I forward you to my source: [Koomri-Text-Segmentation](https://github.com/koomri/text-segmentation)
Please, take into account that I have no connections to Koomri. If you encounter problems with their script, you need to contact them.