The Taskmaster corpus consists of three datasets, Taskmaster-1 (TM-1), Taskmaster-2 (TM-2), and Taskmaster-3 (TM-3), comprising over 55,000 spoken and written task-oriented dialogs in over a dozen domains.
google-research-datasets/Taskmaster
Please see the readme file as well as our 2019 EMNLP paper linked here -->