Slashdot Mirror
George Mason University Speech Accent Archive
JT Olds writes "Apparently George Mason University is running a project to document differences in speech and accents from different backgrounds and the like. They have a paragraph that 306 sample readers have read and recorded, and all of these sound files are categorized by background, gender, age, and other things. They say that this is primarily for teaching and learning, and is especially useful for any linguists out there, but I just thought it was cool. The sound bytes are released under the Creative Commons license. Of course, the Google cache of the main frame is here.
As a side note, I did get the link to this from Penny Arcade's Jerry Holkins."
No cockney support? Insensitive clods...
It's nice they named a University after him, after all, he did save Jack Bauer's life by swapping seats on the plane with the nuke...
We were just talking about how the British English language was the true "natural" English language, all other derived languages that were English with an accent. For example, If I (a person who lives in America and speaks US English; no born American (thank goodness)) were to go to England and converse with an Englishman; who would have the accent, me or him? The obvious answer, as a lot of Americans fail to realize, is me.
Getting speakers of English as a foreign language to repeat a standard English phrase. It's highly unlikely that this produces accents in the sense of two speakers of the same language would recognise. I.e. would a Flemish Dutch speaker recognise the accent of a Dutch speaker from Amsterdam when mangled through an English phrase? Somehow, I don't think so.
It might be useful for tracing people's origins when they are in an Anglosaxon country. But you might as well just ask them.
What would be more useful, perhaps, is a study of the relative differences in accents between native speakers of the "same" language, and how these differences come about.
Ceci n'est pas une signature
This is a really wonderful idea. However, I worry that it has a copule significant problems for researchers. First, for computer analysis work, a paragraph is likely too short to be useful. It can take a *lot* of audio data to make up for one-time variations. Second, cleanliness of the recording. Since anyone can submit a recording, not only will the recording environments and devices differ, but it is unlikely that any recordings will be made in the kind of studio-quality or lab-quality environment that would make these most useful for analysis work.
I'm not a speech synth/recognition researcher, but I do know that generally, for speech research, much stricter constraints are placed on audio being acquired. The extreme variety of the site is nice, but I'm not sure that it outweighs the drawbacks.
May we never see th
When I found this site a few days ago (linked on Penny Arcade), one of the first things that came to mind was how useful it could be to an actor who has to learn how to do a certain accent. In some of the more common accents they even have a list of rules on how most speakers of that other language speak (e.g. many Japanese speakers reverse their R's and L's).
I belong to the ______ generation.
Actors/voice actors have "dialect tapes" which they study to learn accents. I have a few and generally they start by giving vowel substitutions, common phrases and syntax, and then move on to insanely boring phrases you must repeat while trying to copy their accents and inflections.
Religion is a gateway psychosis. -- Dave Foley
I believe the correct answer is both. Everyone has an accent. An accent is the part of speech which is neither specific to an individual or to the language. It varies by region, background, or time period. If you were to go back to the Old English days (there is no "single" English language as it has evolved over time) it is unlikely that anyone would understand you. The same for the Brits.
Maybe the question you meant was which is closer to "correct". If you consider correct to be closer to the root of the evolutionary language tree then the Brittish are probably closer since the Americans' language changed more quickly since the split.
Japanese people don't REVERSE L and R, they just can't pronounce L at all.
This is not particularly correct. Japanese has neither an R nor an L; it has a sound that stands somewhere roughly between the two (whether or not it sounds more like an R or an L depends on the exact speaker, their particular regional accent, and to a certain extent, their gender). And while Japanese speakers of English do not always or even consistently reverse the two consonants, as a consequence of growing up in an environment where the two sounds were conflated, they often have trouble distinguishing the two and have trouble remembering which tongue positioning they should be using for a particular word. Hence it is not uncommon to hear a native Japanese speaker produce an R instead of an L, or vice-versa, in English.
If you want proof of this, just look on any Japanese Katakana or Hirigana chart. These contain all the phonetic sounds in the Japanese language. notice there is no L.
That proves nothing, as Katakana and Hiragana charts contain neither Rs nor Ls; they contain, by definition, Katakana and Hiragana. On an English translation (and the key word here is "translation", as in close approximation of the sounds in english) of the (ra ri ru re ro) portion of the charts they are often presented as R sounds (as this is what they tend to sound like, especially when produced by male speakers in the standard accent), but it is not truly an R (or L sound), as the tounge is at a different position with respect to the upper teeth, and it shares elements in common with the R, L (and to a certain extent D) sounds.
The IDEA archive has a far more complete collection of accents and voice samples. Excellent source material for geeks who work in film, TV or theater.