Identifying Argumentative Questions in Web Search Logs

verfasst von

Yamen Ajjour, Pavel Braslavski, Alexander Bondarenko, Benno Stein

Abstract

We present an approach to identify argumentative questions among web search queries. Argumentative questions ask for reasons to support a certain stance on a controversial topic, such as ''Should marijuana be legalized?'' Controversial topics entail opposing stances, and hence can be supported or opposed by various arguments. Argumentative questions pose a challenge for search engines since they should be answered with both pro and con arguments in order to not bias a user toward a certain stance. To further analyze the problem, we sampled questions about 19 controversial topics from a large Yandex search log and let human annotators label them as one of factual, method, or argumentative. The result is a collection of 39,340 labeled questions, 28% of which are argumentative, demonstrating the need to develop dedicated systems for this type of questions. A comparative analysis of the three question types shows that asking for reasons and predictions are among the most important features of argumentative questions. To demonstrate the feasibility of the classification task, we developed a BERT-based classifier to map questions to the question types, reaching a promising macro-averaged F>sub>1-score of 0.78.

Externe Organisation(en)

Universität Leipzig
Bauhaus-Universität Weimar

Typ

Aufsatz in Konferenzband

Seiten

2393-2399

Anzahl der Seiten

7

Publikationsdatum

07.07.2022

Publikationsstatus

Veröffentlicht

Peer-reviewed

Ja

ASJC Scopus Sachgebiete

Software, Information systems, Computergrafik und computergestütztes Design

Elektronische Version(en)

https://doi.org/10.1145/3477495.3531864 (Zugang: Geschlossen)