2013 Blizzard Challenge
« previous post | next post »
From Simon King:
I am pleased to announce that the English section of this year's Blizzard Challenge listening test is now live. Please help us out by taking part, and encouraging your colleagues, students, friends, contacts, etc. to take part too. It's your chance to hear a range of speech synthesisers, including some really good ones. Please circulate this message widely – for example, on mailing lists, forums and using social media – we need to reach as many people as possible in the coming month or so.
In order to participate, sign up here and follow the instructions.
Eric P Smith said,
May 20, 2013 @ 2:05 pm
Good fun.
This may be idiosyncratic of me and may sound pedantic, but I find the predicates “sounds like a totally different person” and “sounds like exactly the same person” to be totally weird. There are not degrees of being the same person. Could the instruction be worded better in future studies?
leoboiko said,
May 20, 2013 @ 2:32 pm
In my idiolect they could very well sound kinda like the same person, a little bit like the same person, more or less like the same person, almost like the same person, et cetera. In the context of the study, where machines are trying to pass as the same person, I find the modifiers (totally/exactly) to be perfectly natural and intuitive.
Eric P Smith said,
May 20, 2013 @ 2:50 pm
@leoboiko: Yes, but to my mind there is a huge difference between "sounding exactly like the same person" (yes) and "sounding like exactly the same person" (no). The modifiers have to modify the verb and not the NP.
Ellen K. said,
May 20, 2013 @ 9:00 pm
Personally, I find that question, and the answer choices, odd for a different reason. They don't sound like a person at all (for the first 5 questions). How can they sound like the exact same person or a totally different person when they don't sound like a person? If I judge how alike the sample is to the reference sample, I'm not judging whether or not they sound like the same person.
Paul Trembath said,
May 21, 2013 @ 4:35 am
@Ellen K.
I have the same problem, and it's difficult to encode the degree of person-ness along with the degree of sameness on a single scale. If, indeed, that's what we are expected to do.
Faldone said,
May 21, 2013 @ 10:36 am
In terms of sounding like the same person I think we have to consider that some of the unnaturalness in the early sections could be viewed as being caused by transmission problems in the medium. Later sections address the issues of naturalness in several dimensions.
Steve Kass said,
May 22, 2013 @ 5:43 pm
I agree entirely with Ellen, and I said as much in my comments at the end of the survey.