=Paper=
{{Paper
|id=Vol-2921/xpreface
|storemode=property
|title=None
|pdfUrl=https://ceur-ws.org/Vol-2921/xpreface.pdf
|volume=Vol-2921
}}
==None==
<pdf width="1500px">https://ceur-ws.org/Vol-2921/xpreface.pdf</pdf>
<pre>
                                                Preface

Welcome to the first shared task on “Same Side Stance Classification”, collocated with the Sixth
Workshop on Argument Mining at the ACL 2019 in Florence. With the same side classification task
we address the problem of classifying a pair of arguments dealing with the same topic into one of the
two classes (1) same stance (the two arguments have the same stance, i.e., both are ‘pro’ or both are
‘con’) or (2) opposite stance (the two arguments have opposite stance, i..e., one is ‘pro’ and the other
one is ‘con’). In contrast to the related task of stance classification (i.e., classifying a single argument
into ‘pro’ or ‘con’ towards a topic), same side stance classification is simpler, likely to be solvable in
a topic-agnostic fashion. It is probably the most basic stance analysis problem and of high importance
for various applications, such as structuring a discussion or filtering mislabeled arguments in a large
argumentation corpus.

In order to operationalize this task we have constructed a new dataset based on the args.me corpus,
comprising 13,906 arguments from two topics: “abortion” and “gay marriage”. The arguments were
organized to analyze two settings: within- and cross-topics, where the latter is constrained by the fact
that the topics for training and test are different. Nine research groups participated in the shared task.
The groups proposed several systems based on various approaches such as utilizing a Siamese neural
network, vanilla BERT, a multi-task deep network, and bidirectional LSTM. The systems were evaluated
based on their accuracy.

From the eleven submitted systems, we invited the developing groups of six selected systems to submit
papers that describing their solution approach. Each paper was reviewed by at least three reviewers, and
the accepted papers are published in this proceeding. Overall, the acceptance rate is 55% (6 out of 11). In
addition to the accepted papers, the proceeding includes one paper written by the organizers, introducing
the task, elaborating on the task dataset and settings, and summarizing the submitted approaches.
Besides, the proceeding includes one invited paper that analyzes the output of the submitted systems.

Yamen Ajjour
Khalid Al-Khatib
Philipp Cimiano
Roxanne El Baff
Basil Ell
Benno Stein
Henning Wachsmuth

(Same-Side Stance Classification Shared Task 2021 organizers)



We thank all the reviewers in the program committee for their valuable feedback.

Program committee:

   • Yamen Ajjour, Bauhaus-Universität Weimar

   • Khalid Al-Khatib, Leipzig University

   • Milad Alshomary, Paderborn University

   • Alexander Bondarenko, Martin Luther University Halle-Wittenberg

   • Shahbaz Syed, Leipzig University

   • Magdalena Wolska, Bauhaus-Universität Weimar

</pre>