TC-Star Evaluation Information (WP4)

subglobal1 link | subglobal1 link | subglobal1 link | subglobal1 link | subglobal1 link | subglobal1 link | subglobal1 link
subglobal2 link | subglobal2 link | subglobal2 link | subglobal2 link | subglobal2 link | subglobal2 link | subglobal2 link
subglobal3 link | subglobal3 link | subglobal3 link | subglobal3 link | subglobal3 link | subglobal3 link | subglobal3 link
subglobal4 link | subglobal4 link | subglobal4 link | subglobal4 link | subglobal4 link | subglobal4 link | subglobal4 link
subglobal5 link | subglobal5 link | subglobal5 link | subglobal5 link | subglobal5 link | subglobal5 link | subglobal5 link
subglobal6 link | subglobal6 link | subglobal6 link | subglobal6 link | subglobal6 link | subglobal6 link | subglobal6 link
subglobal7 link | subglobal7 link | subglobal7 link | subglobal7 link | subglobal7 link | subglobal7 link | subglobal7 link
subglobal8 link | subglobal8 link | subglobal8 link | subglobal8 link | subglobal8 link | subglobal8 link | subglobal8 link

TTS Evaluation - Run #2

 

TC-STAR Evaluation Run #2 for TTS will take place from Feb 24, 2006 to Mar 31, 2006. The complete schedule can be seen here.

Before the proper evaluation run, participants have access to development data composed of plain text, SSML and XML files, depending on the evaluation tasks. See below development data.

TTS evaluation will be run in 3 languages: English, Spanish and Chinese Mandarin.

Documentation: TC-STAR WP3/ECESS meeting, Berlin January 2006: "TTS Evaluation 2 - Status and Evaluation criteria" : presentation

TTS evaluation data can be downloaded here.

Back to Top

TTS Evaluation Tasks

Modules
Languages
Text Processing UK, CN
Prosody Generation UK, ES, CN
Acoustic Synthesis UK, CN
System UK, ES, CN
Intra-lingual VC UK, ES, CN
Crosslingual VC ES/UK
Expresiv. STL ES
End-to-end UK/ES

 

Evaluation of speech synthesis Modules

M1: Evaluation of text processing

Evaluation of Normalization

Evaluation of Words segmentation (CN)

Evaluation of EOS (End-Of-Sentence) detection (western language)

Evaluation of POS Tagger

Evaluation of grapheme-to-phoneme conversion

M2: Evaluation of prosody

M21-Evaluation of prosody using segmental information

M22-Judgment test using delexicalised utterances

M23-Judgment test using delexicalised utterances

M3: Evaluation of Acoustic synthesis

M31-Functional test (SUS)

M32-Judgment test

Evaluation of Research Topics

Evaluation of voice conversion

Evaluation of intralingual voice conversion

Evaluation of crosslingual voice conversion

Evaluation of VC excluding prosody:
  • Comparison of speaker identities
  • Evaluation of overall speech quality
Evaluation of VC excluding prosody:
  • VC2.1: comparison of speaker identities
  • VC2.2: evaluation of overall speech quality

Evaluation of Expressive speech

ES1-Judgment test

ES2-Functional test (SUS test)

Evaluation of TTS component

Evaluation of the speech synthesis component (whole TTS System)

Evaluation of the synthesis component using ASR output

End-to-end evaluation

Evaluation of speech synthesis within end-to-end (Speech To Speech Translation) system

Back to Top

TTS Participants

Participants/Modules Text processing Prosody Generation Acoustic synthesis TTS component Intralingual VC Crosslingual VC Expressive Speech

End-to-End

ATT       UK + ES        
CAS CN CN   CN CN      
IBM   UK + ES   UK + ES       UK->ES
IBM China       CN UK/CN      
Nokia     UK + CN CN UK      
Uni. Dresden     UK          
Uni. Munich UK              
UPC   UK + ES   UK + ES UK + ES UK-ES ES UK->ES
Siemens UK     UK UK + ES UK-ES    

Back to Top

TTS Resources

Modules
Languages
Corpora
Text Processing
UK
EPPS
CN
863 program data
Prosody Generation
UK
EPPS
ES
EPPS
CN
863 program data
Acoustic Synthesis
UK
EPPS
CN
863 program data
System
UK
EPPS
ES
EPPS
CN
863 program data
Intra-lingual VC
UK
EPPS
ES
EPPS
CN
863 program data
Crosslingual VC
ES/UK
EPPS
Expresiv. STL
ES
EPPS
End-to-end
UK/ES
EPPS

Back to Top

TTS Development Data

Data

Language

Files

DTD

Input of text processing module (M1 module)

Input of prosody module (M2 module)

Input acoustic synthesis module (M3 module)

The specifications for the production of the input for Modules M2 and M3.

UK English data/TTS/Run2/English.zip

DTD

Input of prosody module (M2 module)

Input of acoustic module (M3 module)

Input of TTS component

The specifications for the production of the input for Modules M2 and phoneme segmentation.

Spanish data/TTS/Run2/Spanish.zip

DTD

Input of text processing module eval 1(M1 module)

Input of prosody module (M2 module)

Input of acoustic module (M3 module)

The specifications for the production of the input for Modules M2 and M3.

Chinese Mandarin data/TTS/Run2/Chinese.zip

 

Data

Language + input/output format

DTD UK, ES, CN
Evaluation of text processing module (M1)  
Input and output of M1 UK
Input and output of M1 CN
Evaluation of prosody  
Input of prosody module ES, EN (Siemens M1 output)
Output of prosody module ES: Output of UPC M2 output
Evaluation of acoustic module  
Input of acoustic module ES: Output of UPC M2 output
Output of acoustic module .wav file
Evaluation of TTS system  
Input of TTS system same as input of M1 module
Output of TTS system .wav file
Evaluation of VC  
Input .wav file
Output .wav file
Evaluation of ES  
Labelled speech (WP1 orthographic conventions, starting and ending times for each word)  
End-to-end evaluation  
Output of WP1 system  

Back to Top

TTS Test data

You can find the specifications for the format of the input/output here. Please read carefully this document before running your system !

Modules
Languages
M1 - Text processing evaluation  
Text Normalisation (m1.1), End-Of-Sentence detection (m1.2) , POS tagging (m1.3), Grapheme-to-Phoneme conversion (m1.4) M1-EN

Word Segmentation and POS tagging (m1.2-m1.3)

Grapheme-to-Phoneme conversion (m1.4)

M1.2-M1.3-CN

M1.4-CN

M2 - Evaluation of Prosody Generation module

M2-EN

M2-ES

M2-CN

M3 - Evaluation of Acoustic synthesis module  
M3.1 - Acoustic Synthesis - SUS sentences

M3.1-EN

M3.1-CN

M3.2 - Acoustic Synthesis - EPPS-EN data, 863-CN data

M3.2-EN

M3.2-CN

S - Evaluation of TTS component  
S1- System - EPPS-EN data, EPPS-ES data, 863-CN data

S1-EN

S1-ES

S-CN

S2 - System - ASR + SLT outputs

S2-EN

S2-ES

Evaluation of Voice conversion  
IVC - Intra-lingual VC

IVC-EN (updated with new UKL)

IVC-ES

IVC-CN

CVC - Crosslingual VC ES/UK
ES - Evaluation of Expresiv. STL ExSp-ES
E2E - Evaluation of End-to-end (ASR+SLT+TTS) system UK/ES

Back to Top

TTS Results

!!!!! Information about the results can be found in the README (updated on 2006/04/06 -> VC1 and M31 sections)

!!!!! Information about the evaluation tasks, Language Resources, protocols, metrics and results can be found in the D16 - Evaluation report - TTS part - is available here. (updated on 2006/04/25 with results of 2006/04/06 for EN_VC1, EN_VC2, ES_VC1 and results of Z_S1, Z_S2)

Modules
Results
Raw Results
M1-Text processing evaluation

M1-CN

EN_M1

Chinese (31/03/2006)

English (25/04/2006) Updates: Z_S1, Z_S2, M31: sclite files

Spanish (25/04/2006) Updates: Z_S1, Z_S2, M31: sclite files

Audio file list

Text file list

M21-Evaluation of prosody using segmental information

M22-Judgment test using delexicalised utterances

M23-Judgment test using delexicalised utterances

M32-Judgment test

S1-TTS component (Z_S1 : test with AT&T system)

S2-TTS component (Z_S2 : test with AT&T system)

VC1-Comparison of speaker identities (Intralingual, crosslingual)

VC2- Evaluation of overall speech quality (Intralingual, crosslingual)

ES1-Judgment test

ES2-Functional test

Chinese (31/03/2006)

English (25/04/2006) Updates: EN_VC1, EN_VC2

Spanish (25/04/2006) Updates: ES_VC1

M31 - Functional test (SUS tests)

CN sclite results (31/03/2006)

EN sclite results (31/03/2006)

S2 - Functional test (SUS tests)

Test without AT&T system (S2)

EN sclite results (25/04/2006)

ES sclite results (25/04/2006)

Test with AT&T system (Z_S2)

ES sclite results (25/04/2006)

EN sclite results (25/04/2006)

Back to Top