Dataset · pfb30
Annotated dialogue dataset for chatbot training (MultiWOZ 2.2)
pfb30/multi_woz_v22
Updated multi-domain human dialogue corpus with corrected state and slot annotations.
What's inside
The dataset contains 10,000 to 100,000 examples of written human dialogues in English covering various everyday topics (e.g., restaurant reservations). Version 2.2 fixes dialogue state annotation errors in 17.3 percent of turns compared to version 2.1 and introduces standardized span annotation for slots with high cardinality values.
What you can do with it
- Training dialogue state tracking systems in task-oriented chatbots
- SFT fine-tuning of conversational assistants for reservation and support scenarios
- Training models to extract named entities and slots from user utterances
- Tasks
- Text generation, Fill mask, Token classification, Text classification
- Languages
- English
- Size
- 10K–100K rows
- Downloads this month
- 2.5K
- Likes
- 41
Models trained on it
How to load it
from datasets import load_dataset
ds = load_dataset("pfb30/multi_woz_v22")