takiholadi/kill-me-please-dataset

Dataset Card for Kill-Me-Please Dataset

Dataset Summary

It is an Russian-language dataset containing just over 30k unique stories as written as users of https://killpls.me as of period from March 2009 to October 2022. This resource was blocked by Roskomnadzor so consider text-generation task if you want more stories.

Languages

ru-RU

Dataset Structure

Data Instances

Here is an example of instance:

{'text': 'По глупости удалил всю 10 летнюю базу. Восстановлению не подлежит. Мне конец. КМП!'
 'tags': 'техника'
 'votes': 2914
 'url': 'https://killpls.me/story/616'
 'datetime': '4 июля 2009, 23:20'}

Data Fields

• text: a string containing the body of the story
• tags: a string containing a comma-separated tags in a multi-label setup, fullset of tags (except of one empty-tagged record): внешность, деньги, друзья, здоровье, отношения, работа, разное, родители, секс, семья, техника, учеба
• votes: an integer sum of upvotes/downvotes
• url: a string containing the url where the story was web-scraped from
• datetime: a string containing with the datetime the story was written

Data Splits

The has 2 multi-label stratified splits: train and test.

Dataset Split	Number of Instances in Split
Train	27,321
Test	2,772