---
dataset_info:
  features:
  - name: timestamp
    dtype: int64
  - name: score
    dtype: float64
  - name: subject
    dtype: string
  - name: full_title
    dtype: string
  - name: board_name
    dtype: string
  - name: views
    dtype: int64
  - name: unique_posters
    dtype: int64
  - name: conversation
    list:
    - name: content
      dtype: string
    - name: role
      dtype: string
  splits:
  - name: train
    num_bytes: 806020266.0
    num_examples: 26956
  download_size: 422130168
  dataset_size: 806020266.0
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
---

# 2ch unSafety Dataset


The 2ch unSafety Dataset is a comprehensive collection of text data scraped from the Russian anonymous imageboard 2ch (Двач). This dataset captures the raw, unfiltered discourse found across various boards and threads on 2ch, representing a wide range of topics, discussions, and opinions. Due to the nature of 2ch as an anonymous and uncensored platform, the dataset includes a diverse spectrum of content, ranging from casual banter to controversial, explicit, and potentially offensive material.

Key Features:

	•	Content Variety: The dataset encompasses threads from multiple categories such as technology, politics, relationships, hobbies, gaming, memes, and more, providing a broad view of online discussions.
	•	Anonymity: Since 2ch is an anonymous forum, the dataset lacks any personal identifying information, making it suitable for studying the impact of anonymity on online behavior and communication patterns.
	•	Unfiltered Language: The text is collected in its raw form, which means it includes slang, profanity, hate speech, and other forms of explicit or harmful language, but converted to chat format
	•	Time Coverage: till august 2024
	•	Board Diversity: Data is sourced from a variety of boards (e.g., /b/ - random, /pol/ - politics, /vg/ - video games), providing a rich mixture of thematic discussions.