Hungarian Broadcast News Database

69 Last view: 2021-07-30

Hungarian Broadcast News Database

hu-broadcast

http://alpha.tmit.bme.hu/speech/

ID:

204 The Hungarian Broadcast News (HBN) database was collected as a member of the
Broadcast News Interest Group of COST278, the COST action on Speech and
Language Interaction in Telecommunications in cooperation of 10 different
institutions throughout Europe. The Hungarian material consists of 3h and 30
minutes of recordings, transcribed and annotated (on audio level), using the conventions of NIST
(National Institute of Standards and Technology, USA). The HBN is freely available for non-commercial research purposes, however it is not redistributable and the original video may never be broadcasted or played to public.

You don’t have the permission to edit this resource.

DistributionAvailability

Available - Restricted Use

Start date: 30/11/2011

Licence

MS - NC - No ReD - ND

Restrictions: Academic - Non Commercial Use

Download location: hidden

Distribution Access/Medium: Downloadable

Licensors:

Henk Tamás

Distribution rights holders:

Klára Vicsi

IPR Holder

Klára Vicsi

Contact Person

Klára Vicsi

text
audio

Monolingual text corpusLanguages

Hungarian (25,257 Words)

Linguality

Linguality type: Monolingual

Size

952 Turns

25,257 Words

Character encoding

Mac Dingbat

Creation

Creation mode: Manual

Original Sources

TV broadcast transcription

Monolingual audio corpusLanguages

Hungarian (3.25 Hours)

Linguality

Linguality type: Monolingual

Size

3.25 Hours

Audio duration

194 Minutes

AnnotationSpeech Annotation - Orthographic Transcription

Annotated elements: Background Noise

Segmentation level: Utterance

Format: Transcriber 1.4.2

Annotation Mode: Manual (Segmentated into speaker turns, speaker noises, overlapping music are also marked.)

Annotation Tools:

Transcriber 1.4.2

Start date: 01/05/2005

End date: 30/11/2005

Size: 952 Turns

Content

Speech items: Other

Noise Level: Low

Setting

Naturality: Spontaneous

Conversational type: Multilogue

Scenario: Other

Audience: Large Public

Interactivity: Non Interactive

Audio Formatswave/audio (194 Minutes)

Compression: False

Recording quality: Very High

Quantization: 16

Number of tracks: 1

Sampling rate: 16000

Signal encoding: LinearPCM

Recording

Recording platform software: other

Recording environment: Studio

Recording device type details: Pinnacle DV500

Recording device type: Other

CreationOriginal Sources

TV broadcast (public service news)

Resource Creation

Resource Creator

Klára Vicsi

Creation lasted: 01/12/2004 - 31/12/2005

Metadata

Created: 24/11/2011

Last Updated: 24/11/2011

Source: CESAR

Revision: 2011-11-24

Metadata Creator

György Szaszák

Version

Version: 1.1

Last Updated: 31/12/2005

Usage

Foreseen UseNlp Applications

Use NLP Specific: Discourse Analysis, Speaker Identification, Speech Analysis, Speech To Speech Translation, Speech Understanding

Human UseActual Use - Human Use

Use NLP Specific: Speaker Identification, Speech Analysis, Speech Recognition

Documentation

Janez Zibert, France Mihelic, Jean-Pierre Martens, Hugo Meinedo, Joao Neto, Laura Docio, Carmen Garcia-Mateo, Petr David, Jan Nouza, Matus Pleva, Anton Cizmar, Andrej Zgank, Zdravko Kacic, Csaba Teleki, Klara Vicsi: The COST278 Broadcast News Segmentation and Speaker Clustering Evaluation - Overview, Methodology, Systems, Results. In: Interspeech 2005 - Eurospeech: 9th European Conference on Speech Communication and Technology. Lisboa, Portugal, 2005, ISCA, pp. 629-632.

People who looked at this resource also viewed the following:

Resources from the same creators