Personal tools
You are here: Home Downloads EMIME Bilingual Database The EMIME Bilingual Finnish/English German/English Database

The EMIME Bilingual Finnish/English German/English Database

README which describes the main features of the EMIME Bilingual Database.

[NOTE 1] GF3_ENG_0069.wav and GF3_ENG_0094.wav are corrupt, containing noise only.


 The EMIME Bilingual Finnish/English German/English Database
        Version 1.0 RELEASE September 2010

                Mirjam Wester
       Centre for Speech Technology Research,
           University of Edinburgh


This Database - The EMIME Bilingual Finnish/English German/English Database - is made available under the Open Database License: Any rights in individual contents of the database are licensed under the Database Contents License: A copy of this license is available in this directory: odbl-10.txt.

This file - README_1.0.txt - describes the EMIME Bilingual database. It is a Finnish/English and German/English bilingual database recorded at the University of Edinburgh in 2009/2010 in the context of the EMIME project ( Seven female and seven male speakers of Finnish, German and English were recorded.

0. Notes before you start
1. Directory structure
2. Wave files
   2.1 Microphones
   2.2 Sampling rate
   2.3 Segmentation
   2.4 Naming convention
3. Prompts
   3.1 English
   3.2 Finnish
   3.3 German
4. Extra info - English talkers' accents

0. Notes before you start

For further documentation and when referencing this database in publications, please refer to: M. Wester "The EMIME Bilingual Database", Technical Report EDI-INF-RR-1388, University of Edinburgh, September 2010.(In this directory: EMIME_DATABASE_ACCENTS.pdf)

FM4 is missing 50 Finnish sentences (#5-54)

For further information contact Mirjam Wester,

1. Directory structure
Example of part of the directory structure:


2. Wave files

Two different microphones were used, a close-talking DPA  4035 mounted on the subjects headphones and a Sennheiser MKH 800 p48 microphone placed about 10cm from subject using an omnidirectional pattern. The speech was sampled at 96kHz 24bit depth and stored directly to a computer. These recordings were subsequently downsampled, using Pro-Tools to 22 kHz 16bit.

DPA  4035 is referred to as microphone 0
Sennheiser MKH 800 is referred to as microphone 1

2.2.Sampling rate
The original 96khz files are not included in this release but can be requested from The 22kH files can be found in the directory "downsampled_22kHz".

2.3. Segmentation
Two different prompting methods were used. Some of the talkers read the prompts from a screen which was driven by a semi-automatic prompting script, and others read the prompts from paper. The recordings obtained with the prompting script have been left as they were recorded, these files have somewhat longer pauses at the start and finish of the utterances and may contain mouth smacks.

The recordings obtained by talkers reading from paper have been hand-segmented so leading and trailing silence and mouth smacks etc, have been removed.

Hand-segmented data:
Finnish female: FF2 and FF7
Finnish male: FM4, FM5, FM6 and FM7
German female: GF1, GF2, GF4, GF5, GF6, GF7
German male: all GM1-GM7
English female: all EF1-EF7
English male: all EM1-EM7

2.4. Naming convention
Filename examples:

FM1 = Finnish male 1
ENG = English
0001 = sentence 1
0 = microphone 0 (close-talking)


FF5 = Finnish female 5
ENG = English
0033 = sentence 33
1 = microphone 1 (Sennheiser)

3. Prompts
Three sets of prompts were created; one for each of the languages: English, Finnish and German.
Each set contains:
 * 25 Europarl sentences,
 * 100 news sentences and
 * 20 semantically unpredictable sentences (SUS).

3.1. English
The news sentences for English were taken from the Wall Street Journal 1 corpus comprising 40 enrolment sentences and 60 test set sentences. Enrolment sentences from wsj1/doc/lng_model/adapt.txt/pl_adapt and the other 60 from wsj1/doc/lng_modl/rec_txt/pl_dt_20.nvp

3.2. Finnish
The Finnish news sentences were selected from the Speecon corpus.

There are two different versions of prompts for sentence 117. (The talkers who read from screen had a different version than those that read from paper prompts)

< 117   tampere on ollut perinteisesti hyvä ja edistyksellinen opetusympäristö
> 117   hänellä on tilillään liki kahdeksan kymmentä maaottelua

FM 1,2,3 and FF 1,3,4,5,6 read the "tampere..." version.
FM 4,5,6, 7 and FF 2,7 read the "hänellä" version.

3.3. German
German news sentences were selected from the test set portion of German Globalphone.

4. English talkers' accents

General description of English talkers' accents:
EF1 = Southern English
EF2 = Scottish
EF3 = Northern American
EF4 = New Zealand
EF5 = Scottish
EF6 = Southern English
EF7 = Northern American

EM1 = Scottish
EM2 = Australian
EM3 = Northern American
EM4 = Southern English
EM5 = Scottish
EM6 = Northern American
EM7 = Southern English

M. Wester "The EMIME Bilingual Database", Technical Report, University of Edinburgh, September 2010. contains a description and results of an accent rating experiment in which the degree of foreign accent for all the talkers was rated by native English, Finnish and German listeners.


Document Actions