Following the success of DBDC3, we are please to organize DBDC4 as part of the WOCHAT workshop held in Sicily in conjunction with IWSDS2019.
The task of dialogue breakdown detection is to detect whether the system utterance causes dialogue breakdown (a situation in a dialogue where users cannot proceed with the conversation) in a given dialogue context. The participants of this challenge will develop a dialogue breakdown detector that outputs a dialogue breakdown label (B: breakdown, PB: possible breakdown, or NB:Not a breakdown) and a distribution of these labels. Although we acknowledge that the schedule can be rather tight, we consider DBDC4 to be a perfect opportunity for you to test your dialogue models, especially in terms of contextual understanding and natural human-computer communication.
Task setting of dialogue breakdown detection
Dialogue example annotated with dialogue breakdown labels.
March 29-April 4 April 1-April 4: Formal run
Registered participants will have access to training and dev data. Evaluation data (unlabeld test data) will be distributed just before the formal run to the registered participants. We will distribute both English and Japanese datasets. A registered participant can submit up to five runs for each language. Prospective registrant can have a look at the datasets of DBDC3, which can be found here.
You can have a look at this page to get started on dialogue breakdown detection challenge! This page shows how to use the baseline and evaluate dialogue breakdown detection results.
We are currently looking for sponsors, who will be able to talk during our special session at IWSDS 2019. In addition, sponsors will be credited in all datasets we distribute and advertised in our WOCHAT website.