Twente Nieuws Corpus

registration: 2009/01/21 07:36:41, last modified: 2016/04/04 09:01:19
title
  • Twente Nieuws Corpus
title.abbreviation
  • TwNC
title.alternative
creator
  • researcher Human Media Interaction Group, Department of Electrical Engineering, Mathematics and Computer Science, University of Twente (UT)
subject
  • multifaceted news corpus
subject.linguisticField
  • computational_linguistics
subject.monoMultilingual
  • monolingual
subject.resourceSubject
  • corpus
description
  • The TwNC is a multifaceted Dutch news corpus, comprising about 530 million words of text data and some audio data useful for language model training. The data includes text data from newspaper and magazine articles, and text and audio data from subtitling and autocues/transcripts of broadcast news shows.
description.price *This metadata is only as a guide
  • Contact : hltgroup@cs.utwente.nl
description.language
  • English
description.inputDevice
  • broadcast
description.inputEnvironment
  • broadcast studios
description.speakingStyle
  • read_speech broadcast news
description.speechMode
description.samplingRate
description.additionalData
publisher
  • Human Media Interaction Group, Department of Electrical Engineering, Mathematics and Computer Science, University of Twente (UT)
contributor
  • developer NWO programme CATCH (http://www.nwo.nl/catch)
  • developer STEVIN (http://www.taaluniversum.org/stevin)
  • developer MultimediaN (http://www.multimedian.nl)
contributor.attribute.author.motherTongue
  • native PCM Publishers (for newspaper and magazine articles)
  • native Dutch National Broadcast Foundation (NOS) (for teleprompter files)
contributor.attribute.author.dialect
  • standard_dialect
contributor.attribute.author.level
  • professional
contributor.attribute.author.age
  • adult
contributor.attribute.author.gender
contributor.attribute.speaker.motherTongue
  • native
contributor.attribute.speaker.dialect
  • standard_dialect
contributor.attribute.speaker.level
  • professional
contributor.attribute.speaker.age
  • adult
contributor.attribute.speaker.gender
contributor.attribute.speaker.number
date.created
  • 1998-00-00 2005-00-00
date.issued
date.modified
type
  • Text
  • Sound Recorded broadcast news
type.discourseType
  • report
type.linguisticType
  • primary_text
type.purpose
  • developing_technologies To collect data for the training of language models and acoustic models to be incorporated into a system for large vocabulary speech recognition for Dutch to be deployed in the broadcast news domain and as a baseline system in other domains that lack large amounts of example data
type.style
  • written newspaper articles
  • speech television subtitles
  • written teleprompter (auto-cues) files (the texts the newsreaders read from the teleprompter)
  • written both manually and automatically generated broadcast news transcripts
  • speech broadcast news audio
type.form
  • unfixed
type.sentence
  • short
  • long
type.annotation
  • annotated
type.annotationSample
type.sample
format.extent
  • Total Nuber of Words and Audio Files : approximately 530 million words of text and about 800 files of broadcast news audio
  • Number of Words from Newspapers and Magazines : 547 M words
  • Number of Words from BN Subtitles : 2.3 M words
  • Number of Words from BN Autocues : 3.3 M words
  • Number of Broadcast News Shows Recorded : 828 shows
  • Number of Transcripts for Broadcast News Shows : 26
format.medium
format.encoding
  • UTF-8
format.markup
  • XML
format.functionality
  • the UT Text Normalisation Toolkit (UT-TNT) :
    Partly developed in the MultimediaN project and the STEVIN project SPRAAK.
  • the SHoUT speech recognition toolkit:
    Developed in MultimediaN.
identifier
identifier.doi
identifier.islrn
source
  • NRC Handelsblad and Trouw (1994-1995) (newspaper)
  • Algemeen Dagblad (newspaper)
  • Volkskrant (newspaper)
  • Parool (newspaper)
  • NrcNext (newspaper)
  • Vrij Nederland (magazine)
  • HP De Tijd (magazine)
  • Groene Amsterdammer (magazine)
  • Autocues provided between 1998 and 2005 by the Dutch National Broadcast Foundation (NOS)
language
  • Dutch
language.area
  • Europe
relation
relation.utilization *This metadata is automatically extracted
coverage.temporal
  • 1994 - 2005
coverage.spacial
  • the Netherlands, Europe
rights
  • UT is allowed to redistribute portions of the data under strict licence agreements. Currently, access to the 1999-2002 data can be licensed to individual research groups. The 1994-1995 data was redistributed among the participants of the evaluation campaign within CLEF (Cross-Language Evaluation Forum). Recently, the 1999-2004 data was made available to participants of the Dutch STEVIN speech recognition benchmark evaluation N-BEST, specifically for purposes of language model research.