Comparing BERT with an intent based question answering setup for open-ended questions in the museum domain

BERT-based models achieve state-of-the-art performance for factoid question answering tasks. In this work, we investigate whether a pre-trained BERT model can also perform well for open-ended questions. We set up an online experiment, from which we collected 111 user-generated open-ended questions. These questions were passed to a pre-trained BERT QA model and a dedicated intent recognition based module. We have found that the simple intent based module was around 25% more often correct than the pre-trained BERT model, indicating that open-ended questions still require different solutions compared to factoid questions.

Metadaten
Author:	Md Mahmud Uz-Zaman GND, Stefan Schaffer, Tatjana Scheffler
URN:	urn:nbn:de:bvb:384-opus4-1117180
Frontdoor URL	https://opus.bibliothek.uni-augsburg.de/opus4/111718
URL:	https://www.essv.de/paper.php?id=1125
ISBN:	978-3-959082-27-3OPAC
Parent Title (German):	Elektronische Sprachsignalverarbeitung 2021 (ESSV): Tagungsband der 32. Konferenz, Berlin, 3.-5. März 2021
Publisher:	Förderverein Elektronische Sprachsignalverabeitung e.V.
Place of publication:	Dresden
Editor:	Stefan Hillmann, Benjamin Weiss, Thilo Michael, Sebastian Möller
Type:	Conference Proceeding
Language:	English
Year of first Publication:	2021
Publishing Institution:	Universität Augsburg
Release Date:	2024/02/29
First Page:	247
Last Page:	253
Series:	Studientexte zur Sprachkommunikation ; 99
Institutes:	Philologisch-Historische Fakultät
	Philologisch-Historische Fakultät / Angewandte Computerlinguistik
	Philologisch-Historische Fakultät / Angewandte Computerlinguistik / Lehrstuhl für Angewandte Computerlinguistik (ACoLi)
Dewey Decimal Classification:	0 Informatik, Informationswissenschaft, allgemeine Werke / 00 Informatik, Wissen, Systeme / 004 Datenverarbeitung; Informatik
Licence (German):	Deutsches Urheberrecht

Open Access