The Webis-SameSideAdversarial-21 dataset contains 175 cases. It is manually created to reveal the ability of models to solve different types of adversarial cases for same side stance predictions more systematically. The examples are derived from the dataset used in the Same Side Stance Classification shared task.

We have selected 25 distinct arguments from the "gay marriage" topic that are short and express their stance clearly. For each selected argument, we construct new arguments of four distinct types to obtain two pairs, one with the same stance, and one with an opposing stance:

  • Negation: a simple negation of the argument.
  • Paraphrase: alters important words from the argument to synonymous expressions with the same stance.
  • Argument: uses an argument from the same topic and stance, but semantically completely different regarding the first one.
  • Citation: repeats or summarizes the first argument and then expresses agreement or rejection (a case frequently occurring in the dataset).

The types Paraphrase, Argument, and Citation are also formulated in a negated version to create additional test instances for the opposite stance.


You can access the Webis-SameSideAdversarial-21 corpus on Zenodo.

If you use the dataset in your research, please send us a copy of your publication. We kindly ask you to cite the corpus via [bib]. If you additionally want to link to the dataset, please use the dataset's [doi] for a stable link.


Students: Ahmad Dawar Hakimi.