CGN lexicon

registration: 2008/02/01 08:33:18, last modified: 2008/02/01 08:35:00
title
  • CGN lexicon
title.abbreviation
title.alternative
  • Spoken Dutch Corpus lexicon
creator
  • researcher W. Goedertier, ELIS, Rijks Universiteit Gent (project manager for the entire CGN project)
  • researcher N. Oostdijk, CLS, Katholieke Universiteit Nijmegen (project manager for the entire CGN project)
  • researcher F. Van Eynde, Katholieke Universiteit Leuven (chairman of the working group for corpus annotation)
  • developer Dutch Language Union
  • developer Nijmegen Max Planck Institute (MPI)
subject
  • single-word lexicon
subject.linguisticField
  • lexicography
  • morphology
  • phonetics
  • syntax
  • text_and_corpus_linguistics
subject.monoMultilingual
  • monolingual
subject.resourceSubject
  • dictionary lexicon with informations about the lemma, word form, word sort, syntax, orthographic status, pronunciation, and morphology
description
  • The CGN lexicon is a word list of nearly all the unique word forms in the Spoken Dutch Corpus (Corpus Gesproken Nederlands, CGN), a collection of approximately 900 hours of Standard Dutch from Flemish and Dutch speakers. Each word has been given lexical information about the lemma, word form, word sort, syntax, orthographic status, pronunciation, and morphology. The lexicon only includes word forms written as a single word, not multi-word expressions with spaces.
description.price *This metadata is only as a guide
description.language
  • Dutch
  • English
description.inputDevice
description.inputEnvironment
description.speakingStyle
description.speechMode
description.samplingRate
description.additionalData
publisher
contributor
  • sponsor Netherlands Organization for Scientific Research (NWO) (http://www.nwo.nl/nwohome.nsf/pages/index)
  • sponsor Flemish Government
  • sponsor Dutch Government
  • translator Cumlingua (http://www.cumlingua.com/) (translator)
  • researcher G. Bouma, Rijks Universiteit Groningen (member of working group for the corpus annotation, CGN Project)
  • researcher W. Daelemans, Universiteit Antwerpen (member of working group for the corpus annotation, CGN Project)
  • researcher H. van Halteren, Katholieke Universiteit Nijmegen (member of working group for the corpus annotation, CGN Project)
  • researcher W. Martin, Vrije Universiteit Amsterdam (member of working group for the corpus annotation, CGN Project)
  • researcher M. Moortgat, Rijks Universiteit Utrecht (member of working group for the corpus annotation, CGN Project)
contributor.attribute.author.motherTongue
contributor.attribute.author.dialect
contributor.attribute.author.level
contributor.attribute.author.age
contributor.attribute.author.gender
contributor.attribute.speaker.motherTongue
contributor.attribute.speaker.dialect
contributor.attribute.speaker.level
contributor.attribute.speaker.age
contributor.attribute.speaker.gender
contributor.attribute.speaker.number
date.created
  • 1998-00-00 2003-00-00
date.issued
date.modified
type
  • Text
type.discourseType
type.linguisticType
  • lexicon
type.purpose
  • analysis
  • developing_technologies
type.style
type.form
type.sentence
type.annotation
  • annotated
type.annotationSample
type.sample
format.extent
format.medium
  • On-line use
format.encoding
  • ASCII
format.markup
  • HTML
  • XML
format.functionality
identifier
identifier.doi
identifier.islrn
source
  • CELEX Dutch lexical database - Complete set
  • Reference Database for Dutch (RBN)
  • INL PAROLE Corpus 2004
  • Fonilex pronunciation database
  • Van Dale
  • Word List of the Dutch Language 1995 (Groene Boekje 1995)
  • Eindhoven Corpus
language
  • Dutch
  • Dutch Belgian Dutch (Flemish)
language.area
  • Europe Netherlands and Flanders, Belgium
relation
relation.utilization *This metadata is automatically extracted
  • The accuracy of the part-of-speech tagger TnT trained on CGN is reported to be 96.2% (Oostdijk et al., 2002).
  • The tagger requires an input file with one word per line and was trained on a preliminary version of the CGN corpus (Internal Release 6), and is a purely statistical tagger, based on mono-, bi-, and trigram tag frequencies.
  • The acoustic models needed for this alignment are estimated based on the dictated speech in CGN (a CGN component for which no prosodic annotations are available).
  • In this paper the accuracy of prosodic models based on the CGN database was evaluated and compared to the accuracy of manual annotations.
  • A first conclusion is that the available amount of manual prosodic annotations in the CGN database is sufficient for the development of the proposed baseline prosodic models, a 5 times smaller training database would also suffice.
  • Note that in order to generate those automatic prosodic transcriptions, only an orthographic transcription (available in CGN) is needed.
  • A later update of the CGNcorpus may provide new automatic phonemic transcriptions and segmentations if such ‘matched' 8kHz model provides substantially better results.
  • Our analysis leads us to conclude that in principle CGN data are appropriate for developing a model of the (wh-) questions that people will use in interactive QA.
  • For these words a unique canonical phonetic transcription was obtained, by consulting the general CGN lexicon and these transcriptions were assigned a prior probability of 1.
  • All phonetic transcriptions taken from the lexicon of the Spoken Dutch Corpus were manually verified and checked for consistency with the CELEX entries.
  • We compare the results obtained from the Spoken Dutch Corpus with a similar analysis of a large random collection of FAQs from the internet.
coverage.temporal
  • 1998 - 2003
coverage.spacial
  • Netherlands
  • Flanders, Belgium
rights
  • Copy Right for the entire CGN package: The Dutch Language Union (Nederlandse Taalunie) (http://taalunieversum.org/en/)
  • CGN help desk (tst-cgn@inl.nl)