Subtasks

The ToxHabits task is divided into three different subtasks, each with its own aim. To better understand each of them, please also check the Data description page.

ToxNER (Subtask 1): Substance mention detection

In this subtask, participants have to detect the different instances of substance use in the text. This corresponds to detecting the annotated triggers in a text. This is framed as Named Entity Recognition (NER) task in which participants have to return the corresponding starting and end span (character position) of each annotations and its label.

ToxUse (Subtask 2): Substance use/consumption characteristics detection

In this subtask, participants have to detect the different entities that characterize substance use instances. This corresponds to detecting the annotated arguments in a text. This is framed as Named Entity Recognition (NER) task in which participants have to return the corresponding starting and end span (character position) of each annotations and its label. Due to its special characteristics, the StatusTime argument will not be considered for evaluation in this subtask.

Participation

To participate, please register in Registration, train and test datasets can be found there when published.