Skip to content

Dataset · convai-challenge

Human-bot dialogues with quality ratings

convai-challenge/conv_ai_2

English human-bot conversations annotated with quality ratings for individual turns and full dialogues.

What's inside

It contains English human-chatbot dialogues annotated for quality. The size ranges from 1k to 10k records, with quality ratings provided for both individual utterances and entire conversations.

What you can do with it

  • Training a metric or reward model for automated dialogue system evaluation
  • Selecting the best responses during response generation in conversational chatbots
Tasks
Text classification
Languages
English
Size
1K–10K rows
Downloads this month
758
Likes
41

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("convai-challenge/conv_ai_2")