Spoken corpora

Spoken corpora are corpora that consist of spoken data or material based on spoken data.

Boarnsterhim Corpus (BHC) (Currently unavailable)

The Boarnsterhim Corpus consists of 250 hours of speech in both West Frisian and Dutch by the same sample of bilingual speakers. The corpus contains original recordings from 1982-1984 and a replication study recorded 35 years later. The data collection spans speech of four generations, and combines panel and trend data. ##This corpus is temporarily unavailable because because it is under revision. A new version is expected in the summer of 2024. For more information, please contact Hans van de Velde (HvandeVelde@fryske-akademy.nl) or Wilbert Heeringa, datamanager of the Fryske Akademy (wheeringa@fryske-akademy.nl).##

• 42.6 MB
• version 1.0 (2020)
• data set from 1982-1984 + replication 35 years later
• Download page

COPAS: Corpus Pathologische en Normale Spraak

A collection recordings of almost 200 speakers with an audible speech impediment and a control group of 122 speakers.

• Belgian Dutch
• Download page
• Project page

Corpus Gesproken Nederlands

The Spoken Dutch Corpus contains almost 9 million words of contemporary spoken Dutch from native speakers in Flanders and the Netherlands.

The speech recordings are aligned with several transcriptions (e.g. orthographic, phonetic) and annotations (syntax, POS-tags). Metadata, lexica, frequency lists and the tool Corex which can be used to explore the data are included.

• 900 hours of spoken Dutch
• 1998 - 2004
• tagged, lemmatized, annotated (orthographic/phonetic)
• corpus exploration software (Corex)
• version 2.0.3.
• Project website
• Download page
• Online search with OpenSonar. If you go to Extended Mode you can select to exclusively search in the Corpus Spoken Dutch. (See Corpus querying for more information on OpenSonar.)

IFA Spoken Language Corpus

The IFA Spoken Language corpus is a free (GPL) database of hand-segmented Dutch speech at the phoneme level. It was constructed with off-the-shelf software using speech from 8 speakers in a variety of speaking styles. For a total of 50,000 words (41 minutes/speaker), speech acquisition and preparation took around 3 person-weeks per speaker.

• version 1.0 (2001)
• 4.6 MB
• Download page
• Project website

JASMIN-spraakcorpus

A corpus of contemporary Dutch (Dutch/Flemish) as spoken by children of different age groups, elderly people and non-natives with different mother tongues, and human-machine interaction

• 115 hours of spoken Dutch
• speech of children, elderly people and non-natives, and human-machine interaction
• verbatim transcription, a transcription of the human-machine interaction (HMI) phenomena, POS tagging of the words, and an automatic phonetic transcription
• version 1.0 (2008)
• Recording Speech of Children, Non-Natives and Elderly People for HLT Applications: the JASMIN-CGN Corpus (LREC Proceedings 2008)
• Download page

SABeD -- Spoken Academic Belgian Dutch

The SABeD corpus collection project has started on the 1st of March 2021 and is not yet available. The corpus of spoken academic Belgian Dutch will consist of at least 200 lectures.

• Project website

AUTONOMATA-namencorpus

The AUTONOMATA Spoken Names Corpus is a database with in total about 5000 read first names, surnames, straat names, city names and check words.

• version 1.0 (2008)
• Paper
• Download page

AUTONOMATA-POI-corpus

The AUTONOMATA POI Corpus is a corpus of 800 pronounced points of interest from the Netherlands and Belgium containing names of restaurants, camping sites, cafés, etc.

• Documentation
• Project website
• Download page

Children's Oral Reading Corpus (CHOREC)

The CHOREC Corpus contains recorded, transcribed and annotated read speech (42 GB or 130 hours) of 400 Dutch speaking elementary school children with or without reading difficulties. Analyses of inter- and intra-annotator agreement are carried out in order to investigate the consistency with which reading errors are detected, orthographic and phonetic transcriptions are made, and reading errors and reading strategies are labeled.

• Paper
• Project page
• Download page

BLISS Dialogue Summaries

This dataset consists of Dutch recordings of participants talking with the BLISS dialogue system about their everyday occupations and their favorite activities. The corpus contains 55 recordings with an average duration of 2 minutes and 34 seconds.

• Project page
• Download page

The Ernestus Corpus of Spontaneous Dutch

The Ernestus Corpus of Spontaneous Dutch contains high quality recordings of 10 conversations, each 90 minutes long, between friends or direct colleagues. The corpus was recorded between autumn 1995 and spring 1996 at the Institute of Phonetics of the University of Amsterdam. Professional transcribers have created an orthographic transcription of the corpus by hand, while a phonemic transcription has been created automatically.

• Publication: M. Ernestus (2000). Voice assimilation and segment reduction in casual Dutch: A corpus-based study of the phonology-phonetic interface. Holland Institute of Generative Linguistics, Utrecht.
• Website: [1]
• Mirjam Ernestus (2000). Item "The Ernestus Corpus of Spontaneous Dutch" in collection "Nijmegen corpora of casual speech". The Language Archive. https://hdl.handle.net/1839/a8025f06-cf20-4183-aae5-7c3309bc8c9d