SemEval-2018 task 3 - Irony detection in English tweets Forum

Go back to competition Back to thread list Post in this thread

> constrained vs unconstrained submition

What exactly is considered a constrained submission? Can we use e.g. lexicons, embeddings, clusteres, sentiment labels etc. ?

Thanks for answer.

Posted by: tigi @ Jan. 17, 2018, 1:49 p.m.

Hi

This has been addressed several times in the Google group of the task as well. Constrained systems are *only* trained on the provided data for this task, whereas for unconstrained systems additional training data (e.g. collected and annotated by the participants themselves) were used. The constrained/unconstrained distinction only applies to the training data, not to data used for feature engineering, i.e. constrained systems can make use of additional resources like lexicons, dictionaries, embeddings...

Kind regards
Cynthia

Posted by: cyvhee @ Jan. 22, 2018, 1 p.m.
Post in this thread