This database was created by Nordic Language Technology for the development of automatic speech recognition and dictation in Norwegian. In this version (from 2022), the organization of the data has been altered to make the database more user friendly.
The data set was reviewed again in December 2023; duplicates were removed, metadata was cleaned up and filenames were standardized. See the documentation file for details.
In the original version of the material, the files were organized in a specific folder structure where the folder names were meaningful. However, the file names were not meaningful, and there were also cases of files with identical names in different folders. This proved to be impractical, since users had to keep the original folder structure in order to use the data.
The files have been renamed, such that each file name is unique and meaningful regardless of the folder structure. The original metadata files were in spl format; these have been converted to JSON format. The metadata files are anonymized, and the text encoding has been converted from ANSI to UTF-8. Metadata and transcriptions are also available as CSV files.
See the documentation file for a full description of the data and the changes that have been made to the database.
This database was created by Nordic Language Technology for the development of automatic speech recognition and dictation in Norwegian. In this version (from 2022), the organization of the data has been altered to make the database more user friendly.
The data set was reviewed again in December 2023; duplicates were removed, metadata was cleaned up and filenames were standardized. See the documentation file for details.
In the original version of the material, the files were organized in a specific folder structure where the folder names were meaningful. However, the file names were not meaningful, and there were also cases of files with identical names in different folders. This proved to be impractical, since users had to keep the original folder structure in order to use the data.
The files have been renamed, such that each file name is unique and meaningful regardless of the folder structure. The original metadata files were in spl format; these have been converted to JSON format. The metadata files are anonymized, and the text encoding has been converted from ANSI to UTF-8. Metadata and transcriptions are also available as CSV files.
See the documentation file for a full description of the data and the changes that have been made to the database.
Extended metadata
resource Common Info:
resource Type: corpus
identification Info:
resource Name: NST norsk ATG-database (16 kHz) – reorganisert
resource Name: NST Norwegian ASR Database (16 kHz) – Reorganized
description: Denne databasen er laget av Nordisk språkteknologi som datagrunnlag for talegjenkjenning og diktering på norsk. I denne versjonen (fra 2022) er dataene strukturert på nytt, slik at databasen enklere kommer til anvendelse.
Datasettet ble gjennomgått på nytt i desember 2023; duplikater ble fjernet, det ble ryddet i metadataene og filnavn ble standardisert. Se dokumentasjonsfilen for detaljer.
I den opprinnelige versjonen av materialet var filene organisert i en bestemt mappestruktur der mappenavnene var meningsbærende. Filnavnene var imidlertid ikke meningsbærende, og det fantes også tilfeller av filer med samme navn i forskjellige mapper. Dette viste seg å være upraktisk, siden brukerne måtte beholde den opprinnelige mappestrukturen for å kunne benytte seg av dataene.
Filene har fått nye navn, slik at alle filnavn er unike og meningsbærende uavhengig av mappestrukturen. De opprinnelige metadatafilene var i spl-format; disse er konvertert til JSON-format. Metadatafilene er anonymisert, og tekstkodingen er endret fra ANSI til UTF-8. Metadata og transkripsjoner foreligger også som CSV-filer.
Se dokumentasjonsfilen for en full beskrivelse av dataene og endringene som er gjort.
description: This database was created by Nordic Language Technology for the development of automatic speech recognition and dictation in Norwegian. In this version (from 2022), the organization of the data has been altered to make the database more user friendly.
The data set was reviewed again in December 2023; duplicates were removed, metadata was cleaned up and filenames were standardized. See the documentation file for details.
In the original version of the material, the files were organized in a specific folder structure where the folder names were meaningful. However, the file names were not meaningful, and there were also cases of files with identical names in different folders. This proved to be impractical, since users had to keep the original folder structure in order to use the data.
The files have been renamed, such that each file name is unique and meaningful regardless of the folder structure. The original metadata files were in spl format; these have been converted to JSON format. The metadata files are anonymized, and the text encoding has been converted from ANSI to UTF-8. Metadata and transcriptions are also available as CSV files.
See the documentation file for a full description of the data and the changes that have been made to the database.
This database was created by Nordic Language Technology for the development of automatic speech recognition and dictation in Norwegian. In this version (from 2022), the organization of the data has been altered to make the database more user friendly.
The data set was reviewed again in December 2023; duplicates were removed, metadata was cleaned up and filenames were standardized. See the documentation file for details.
In the original version of the material, the files were organized in a specific folder structure where the folder names were meaningful. However, the file names were not meaningful, and there were also cases of files with identical names in different folders. This proved to be impractical, since users had to keep the original folder structure in order to use the data.
The files have been renamed, such that each file name is unique and meaningful regardless of the folder structure. The original metadata files were in spl format; these have been converted to JSON format. The metadata files are anonymized, and the text encoding has been converted from ANSI to UTF-8. Metadata and transcriptions are also available as CSV files.
See the documentation file for a full description of the data and the changes that have been made to the database.