Img

Workshops and Session Series on Chatbots and Conversational Agents

Shared Task Datasets

Thank you for participating in WOCHAT! You can download from this page all data and annotations generated in the shared task. Please be reminded that this is a continuous effort. We appreciate your help with promoting the tasks and engaging new participants.

New participants are encouraged to register by filling up this form. The provided contact information will be only used for updating about any changes related to this data collection activity.

The WOCHAT Shared Task datasets are provided in the table below. All files follow the official shared task XML format, as described in the Annotation Guidelines. The datasets provided in the table include all chatting sessions and annotations collected in both RE-WOCHAT and WOCHAT workshops. (Click on the datafile names in the table to download the corresponding dataset)

Chatbot Sessions Turns Annotations
Datafile Language Overall Annotated Overall Annotated Overall Expert Crowd Others
TickTock.zip English 206 206 5462 2891 3736 - 1005 2731^
IRIS.zip English 163 56 5687 1526 4466 3251 1215 -
Joker.zip English 112 27 4478 1058 2132 1010 1122 -
Sammy.zip English 32 7 1026 171 513 - 513 -
Joker2.zip English 26 - 624 - - - - -
Sarah.zip English 24 - 1108 - - - - -
others.zip* English 19 - 770 - - - - -
Politician.zip Czech 17 - 559 - - - - -
Sammy.zip French 12 - 338 - - - - -
pyEliza.zip English 10 7 324 233 466 466 - -
Sammy.zip Italian 3 - 64 - - - - -
Totals 624 303 20440 5879 11313 4727 3855 2731^


^ Annotations generated by the same users interacting with the chatbot (only chatbot turns are annotated).
* Includes chatting sessions with previous versions of IRIS.