Evaluation: Difference between revisions
mNo edit summary |
No edit summary |
||
| Line 5: | Line 5: | ||
* Assessing the overall quality of a TTS system does not often give interesting information on how to improve the system, since the output os the result of several complex and intermixed processes. | * Assessing the overall quality of a TTS system does not often give interesting information on how to improve the system, since the output os the result of several complex and intermixed processes. | ||
It is generally agreed that the developement of free software can boost assessment and improvement of technologies. | It is generally agreed that the developement of free software can boost assessment and improvement of technologies. As far as speech synthesis is concerned, the community has made several important contributions with: | ||
* [http://www.cstr.ed.ac.uk/projects/festival/ Festival] and more recently Festival 2, which includes a unit-selection synthesizer) | |||
* [http://festvox.org/ FestVox], an extension of Festival with a special focus on voice corpora developements | |||
* [http://tcts.fpms.ac.be/synthesis/mbrola.html MBROLA] a free, multilingual, diphone-based synthesizer. | |||
* [http://freetts.sourceforge.net/docs/ FreeTTS] provides free speech synthesis in JAVA. | |||
* [http://mary.dfki.de OpenMary], a multi-lingual (German, English, Tibetan) and multi-platform (Windows, Linux, MacOs X and Solaris) speech synthesis system. | |||
Developing widely available common datasets is also a primary importance for encouraging informative comparative tests of synthesis techniques. For American English, the CMU ARCTIC dataset available in the framework of the [http://www.festvox.org/blizzard/blizzard2005.html BLIZZARD challenge] is an example to follow (and adapt to other languages). | |||
There is no universally accepted assessment technique for TTS. In the [http://www.festvox.org/blizzard/bc2005/IS052023.PDF Blizzard challenge], the naturalness of speech synthesizers was judged on the basis of MOS (Mean Opinion Score) tests, while their intelligibility were measured by the WER (word error rate) otbained in two test conditions : a MRT test (modified rhyme test) and a SUS test (using semantically unpredictable sentences). | |||
==HLT-evaluation.org== | ==HLT-evaluation.org== | ||
Another source of information on Speech Synthesis evaluation is the [http://www.hlt-evaluation.org/article.php3?id_article=16 TTS page] in the [http://www.hlt-evaluation.org Human Language Technologies Evaluation] web site. | Another source of information on Speech Synthesis evaluation is the [http://www.hlt-evaluation.org/article.php3?id_article=16 TTS page] in the [http://www.hlt-evaluation.org Human Language Technologies Evaluation] web site. | ||
Revision as of 11:48, 16 May 2006
Towards freely usable software for assesment purposes
Assessing speech synthesis is not as easy as assessing speech recognition, for various reasons:
- Various criteria can be used (do we assess speech intelligibility, or speech naturalness, or the efficiency of the speech component in a given application, etc.).
- It systematically requires subjective tests by human listeners, which makes assessment a heavy task.
- Assessing the overall quality of a TTS system does not often give interesting information on how to improve the system, since the output os the result of several complex and intermixed processes.
It is generally agreed that the developement of free software can boost assessment and improvement of technologies. As far as speech synthesis is concerned, the community has made several important contributions with:
- Festival and more recently Festival 2, which includes a unit-selection synthesizer)
- FestVox, an extension of Festival with a special focus on voice corpora developements
- MBROLA a free, multilingual, diphone-based synthesizer.
- FreeTTS provides free speech synthesis in JAVA.
- OpenMary, a multi-lingual (German, English, Tibetan) and multi-platform (Windows, Linux, MacOs X and Solaris) speech synthesis system.
Developing widely available common datasets is also a primary importance for encouraging informative comparative tests of synthesis techniques. For American English, the CMU ARCTIC dataset available in the framework of the BLIZZARD challenge is an example to follow (and adapt to other languages).
There is no universally accepted assessment technique for TTS. In the Blizzard challenge, the naturalness of speech synthesizers was judged on the basis of MOS (Mean Opinion Score) tests, while their intelligibility were measured by the WER (word error rate) otbained in two test conditions : a MRT test (modified rhyme test) and a SUS test (using semantically unpredictable sentences).
HLT-evaluation.org
Another source of information on Speech Synthesis evaluation is the TTS page in the Human Language Technologies Evaluation web site.
