ELRA - ELRA-U-S 0172 : Verbmobil Data

You are here » Universal Catalogue » Spoken Resources » Telephone

Language Resources

Search Catalogue

Send us information

Would you like to collaborate ?
Contact Us

Languages

Catalog Reference : ELRA-U-S 0172

Verbmobil Data

The aim of the Verbmobil project was the development of a mobile translation system for the translation of spontaneous speech in face-to-face situations. In this perspective, spontaneous speech data had to be collected and transcribed for the training and testing of Verbmobil systems. The different languages concerned are German, English and Japanese.

3,200 dialogs were collected from 1,658 speakers using a close microphone, a room microphone and a telephone:
- 1,454 dialogs for German, 726 for English and 1,020 for Japanese.
Some dialogs were also annotated using a hierachy of 32 dialog acts.
- 79,562 turns: 41,512 for German, 16,104 for English and 21,946 for Japanese.
- 1,520,000 running words: 670,000 for German, 270,000 for English and 580,000 for Japanese.
- 181,6 hours: 96,1 for German, 37,9 for English and 47,7 for Japanese.

A treebank with 85,000 entries was also developed (German: 35,000; English: 30,000; Japanese: 20,000). It was annotated with part-of-speech tags, phrasal categories, grammatical functions and root labels. The treebank was used for the training of the statistical parser, the chunk parser and the development of semantic construction rules and translation transfer rules.

Production

Project : Verbmobil Project

Applications

	Applications possible : Speech recognition#Automatic speech recognition
application Area : Research

Contents

Click on the arrow to display content.

written corpus
Number of languages : Multilingual
Language(s) : German ; Japanese ; English
Annotation Coverage : Partial
Annotation Granularity : Word
Annotation level : Syntactic
speech corpus
Language(s) : German ; English ; Japanese
Source Channel : Microphone#Telephone