# Real Human Conversations Multilingual Dataset

A unified dataset of real human conversations from across the world.

## Languages
- English: Reddit, Discord, Twitch, therapy, Usenet
- Russian: Telegram chats
- Italian: Usenet newsgroups
- Japanese: Speech transcriptions
- Korean: Everyday conversations

## Format (Parquet)
| Column | Type | Description |
| text | string | Conversation text |
| source | string | Origin dataset |
| language | string | ISO code |
| domain | string | Conversation type |
| turns | int | Dialogue turns |
| metadata | string | Extra JSON info |

## Sources
- mookiezi/Discord-Dialogues
- HuggingFaceGECLM/REDDIT_comments
- fsteig/conversations-30gb
- SocialGrep/one-million-reddit-confessions
- Den4ikAI/russian_dialogues_2
- mii-community/UsenetArchiveIT-conversations
- japanese-asr/whisper_transcriptions.reazon_speech_all
- jojo0217/korean_safe_conversation
- lparkourer10/twitch_chat
- Amod/mental_health_counseling_conversations

## Usage
```python
from datasets import load_dataset
ds = load_dataset("asdf98/human-chats-worldwide", split="train")
```