The best performance is currently owned by a native French speaker expert in speech processing, with a rate of 77% correct identification. Non French speakers perform surprisingly well, especially musicians who performed a substantial rate of 62% correct identification. --- Last update: 2011, january 11th, 20.30.
Thank you for participating ! This experiment won't take you more than 15 minutes .
This study aims at evaluating speaking style adaptation in speech synthesis.
Each speaker has his own speaking style which constitutes his voice signature, and a part of his identity. Nevertheless, a speaker continuously adapt his speaking style according to specific communication situations.
Or, each situational context determines a specific mode of production associated with it - a genre - which is defined by a set of conventions of form and content that is shared among all of its productions. In particular, a specific discourse genre relate to a specific speaking style. .
You are going to listen to some speech samples that have been filtered in order to focus on their prosodic dimension.
For each sample, you have to associate it with one of the proposed speaking style as described bellow.
There are 4 four speaking styles for that experiment.
Here is a description and the symbols related to the different speaking styles.
In order to kindly remind you how sound the speaking styles to identify, and to test your identification ability, please start the experiment with the following real speech samples. You can use this pre-experiment to familiarize with the experiment procedure.
Here is a couple of synthesized utterances of the original speaker to be adapted, namely the French actor André Dussolier.
Now, the speaking style of the original speaker has been transformed torwards the different speaking styles. For each utterance, you have to indentify the speaking style in which the original speaker has been transformed. The text is neutral, so you must not consider its content to make your decision.
We strongly recommend you to process the experiment in several steps:
Listening conditions headphones other
Langage french native french speaking other
Expertise (are you familiar to a domain related to speech, such as speech processing, linguistic, ... ?) yes no
Expertise Domain (if you are expert, what is your precise domain: speech, linguistic, music, ...)
Age
Mail address and name (strongly recommanded if you've participated to the previous experiment)
Comments
Before sending please check that all the files of the main experiment got an answer.
Click on "send your answers" to get your result !! You can redo the test as many time as you want to improve your score.
All recordings are Ircam's property.
Public
Experiments
login
edit SideBar