Sprekend Nederland: A heterogeneous speech data collection

David A. Van Leeuwen, Frans Hinskens, Borja Martinovic, Arjan Van Hessen, Stef Grondelaers, Rosemary Orr

Research output: Contribution to journalConference articleAcademicpeer-review

7 Citations (Scopus)

Abstract

Sprekend Nederland is a large-scale efiort to document the variability of Dutch as spoken in the Netherlands anno 2016. A smartphone app was created to record the speech of as many speakers of Dutch as possible, as well as their attitudes (perceptions and evaluations) towards other participants's speech. Initiated by the national broadcast organisation NTR, Sprekend Nederland relies on both traditional and modern media to recruit participants. At this point, about halfway through the project, over 7000 participants have recorded over 200 000 utterances, totalling about 375 hours of speech data, and over a million of attitude judgements have been given. In this paper we report the design and implementation of the data collection, we present some preliminary statistics and demographics, and we outline a number of research possibilities that this data collection ofiers.

Original languageEnglish
Pages (from-to)21-38
Number of pages18
JournalComputational linguistics in the Netherlands journal
Volume6
Publication statusPublished - 1 Dec 2016
Event26th Meeting of Computational Linguistics in the Netherlands, CLIN 2016 - Amsterdam, Netherlands
Duration: 18 Dec 201618 Dec 2016
Conference number: 26

Fingerprint

Dive into the research topics of 'Sprekend Nederland: A heterogeneous speech data collection'. Together they form a unique fingerprint.

Cite this