Overview of Touché 2020

Argument Retrieval

verfasst von

Alexander Bondarenko, Maik Fröbe, Meriem Beloucif, Lukas Gienapp, Yamen Ajjour, Alexander Panchenko, Chris Biemann, Benno Stein, Henning Wachsmuth, Martin Potthast, Matthias Hagen

Abstract

Argumentation is essential for opinion formation when it comes to debating on socially important topics as well as when making everyday personal decisions. The web provides an enormous source of argumentative texts, where well-reasoned argumentations are mixed with biased, faked, and populist ones. The research direction of developing argument retrieval technologies thus focuses not only retrieving relevant arguments for some argumentative information need, but also on retrieving arguments of a high quality. In this overview of the first shared task on argument retrieval at the CLEF 2020 Touché lab, we survey and evaluate 41 approaches submitted by 17 participating teams for two tasks: (1) retrieval of arguments on socially important topics, and (2) retrieval of arguments on everyday personal decisions. The most effective approaches submitted share some common techniques, such as query expansion, and taking argument quality into account. Still, the evaluation results show that only few of the submitted approaches (slightly) improve upon relatively simple argumentation-agnostic baselines—indicating that argument retrieval is in its infancy and meriting further research into this direction.

Externe Organisation(en)

Martin-Luther-Universität Halle-Wittenberg
Universität Hamburg
Universität Leipzig
Skolkovo Institute of Science and Technology
Bauhaus-Universität Weimar
Universität Paderborn

Typ

Konferenzaufsatz in Fachzeitschrift

Journal

CEUR Workshop Proceedings

Band

2696

Anzahl der Seiten

22

ISSN

1613-0073

Publikationsdatum

2020

Publikationsstatus

Veröffentlicht

Peer-reviewed

Ja

ASJC Scopus Sachgebiete

Informatik (insg.)

Elektronische Version(en)

https://ceur-ws.org/Vol-2696/paper_261.pdf (Zugang: Offen)