Dataset for 'Learning Low-Resource End-To-End Goal-Oriented Dialog for Fast and Reliable System Deployment' ACL 2020

'extended-bAbI' is a simulated multi-domain e2e dialog dataset
There are 7 domains: restaurant, flights, hotels, movies, music, tourism and weather
Each domain has train/dev/test dataset (1500/500/1000)


'multiWOZ' is a bAbI-formatted e2e dialog dataset transformed from original MultiWOZ2.1 (only single domain dialogs)
There are 7 domains: attraction, hospital, hotels, police, restaurant, taxi and train
we use delexiconalization for all slot-values, and also normalize the sys_act to reduce the number of candidate responses.


For any information, please feel free to contact Yinpei Dai : yinpei.dyp (at) alibaba-inc (dot) com .

