full-fledged pidgin

The Internet (yes, it’s capitalized, the atechnical pseudo-nerds over at Wired can bite me) has turned up the following amazing image:

This would be a more endearing compliment if the bear were to employ the continuative ‘be’ — implying that the recipient was habitually fine. Still, though, even an ephemeral admission of fineness, when printed on a silk-look pillow and delivered by a wide-eyed stuffed toy, must turn the young ladies’ disinterest to passion.

I started writing a long post analyzing the humorous impact of this low-quality stuffed bear (the sort one might win at a carnival or boardwalk) controlling this gruff, urban vernacular but, really, the bear speak for hisself.

wheel of fortune

I’m currently looking for someone who might be interested in working with me on a speech synthesis research project. The gig would involve some time in the recording studio recording transcripts intended to elicit as-nearly-complete-as-possible a selection of English speech sounds in different contexts, different emotional conditions, and at different paces. Speaker should be comfortable reading copy into a mic (acting experience/training a huge plus), interested in learning something about how his/her vocal tract works, and willing to participate in a number of articulatory measurements (think tubes, electrodes, ultrasound wand placed submentally (beneath your jaw)) for surprisingly little compensation. :)

There would be two sessions:

1. Session 1: high-quality sound recording as described above.
2. Session 2: recording of same transcripts in a phonetics laboratory with various instrumentation installed and attached.

When we’re done you get an open-source speech synthesizer that talks in your voice. Please comment or e-mail me if you’re interested (or know someone who might be).

there’s only a 16% chance that these titles were arranged in this way by pure chance — and that assumes that they all arrived at the theater on the same date. Other arrangements are interpretable, but nothing else is this funny. “knocked up shrek pirates” is still pretty good, but it’s only an NP (people who pirate copies of shrek and have been impregnated).

As an aside, why in the world is WordPress putting a definite article in date strings? Perhaps I can write a paper: The English Date String in its Sentential Aspect.

I’ll bet my advisor could help with that. ;)

(I’m finishing these projects out of order (wavesurfer took me longer to deal with than Praat or Audacity, so the two WaveSurfer posts have languished on the largely-completed-but-unpublished stack for a while now). Exercise 5 is below. Anyone who might be reading this and who is not, in fact, one of the professors for LSA 317: Experimental Phonology should probably move along. I’m not going to provide enough context for this to make sense, sorry. )

I can definitely see that wavesurfer would be a an incredibly useful application if it worked well. As it is, though, I’ve found it wildly frustrating (even on a mac).

(more…)

Every time I measure speech sounds for some project or to answer some question I find myself wishing that the LSA web site (or some similar organization — ASA?) had a few pages devoted to the nuts and bolts of measuring speech sounds: what trade-offs people make, how to make sure you’re being consistent, etc. Then Keith Johnson, I think jokingly, suggested we, the students in his LSA 07 Experimental Phonology class, should post on our blogs about how we take measurements with illustrative screenshots. I’m pretty sure he was kidding, so I’ve decided to do it.

Disclaimer: I believe that I’m following what Pam Beddor taught me, but any terrible decisions or ridiculous errors are entirely my own.

(more…)

I’m posting this exercise out of order because the other two are in states of not-yet-quite-completion. Maybe when they’re done and posted I’ll figure out how to resort the main page.

A. Guinea pigging

toons --> twos

(more…)

this is just a reminder to myself to investigate more about this when I have time. How would a direct realist (or even a motor theorist) respond to a sound change where V becomes nasalized after a voiceless fricative (in the absence of any nasal consonant) due to rhinoglottophilia? Does L&M talk about any such changes?

Come to think of it, the development of tones from voiced/voiceless stops presents a very similar problem to these perception theories.

(this is the last of the LSA 317 exercises I’ll be posting here. the post date/time should reflect when I started working on this exercise, but I didn’t actually end up finishing it until late night on the last night of the institute.)

(more…)

Question 1

What does “skunk” sound like with /s/ gone? How do you account for this
transformation?

/skʌŋk/ with the /s/ removed now sounds like [gʌŋk]. This is because the so-called `voiced’ velar plosive in (at least American) English is actually a voiceless, unaspirated velar plosive with a lower fundamental frequency.

Question 2

What does “damp” sound like with the /s/ in front of it? How do you account for
this transformation?

/s/ + /damp/ now sounds fairly convincingly like [stæmp] (convincingly enough to fool my linguist roommates). I suspect two possible explanations for this phenomenon:

1. English phonotactics prevent the listener from hearing /d/ even though it is acoustically distinct from /t/.
2. The expected /d/ and /t/ sounds are, as described above, more similar than they are dissimilar and the listener uses context to distinguish the two.

These are surprisingly difficult to tease apart with the data I have available. One approach might be just to collect a few hundred /d/ and /t/ productions in various contexts, measure them, and then quantify the similarity or dissimilarity of the two sounds for the speaker. As it is, though, I don’t even have one example of a /t/ for this speaker. Another approach might be to record a speaker saying the minimal pair /tæmp/ and /dæmp/ (in a frame like ’say /tæmp/ again’ and with some distractor words so he doesn’t artificially enunciate them). Again, though, I feel I should be able to make this distinction with the available data.

A third approach might be to move some other sound in front of the /dæmp/ to see if another sound (e.g. /æ/ also causes the /d/ to sound like a [t]. I did this and it does not, for me, make /d/ sound like [t], but this also isn’t a real word so perhaps I’m not listening to it as speech.

I think there’s a clue to this puzzle in the fact that moving the /s/ too close to the /d/ (anywhere from 0 to about 40ms in my informal testing) fails to cause the changed perception — /d/ still sounds like /d/. This /d/ does not differ from a /t/ in terms of voicing and the solution isn’t one or the other of my two stated possibilities but the unification of the two. This perceptual shift is the same as /kʌŋk/ –> /gʌŋk/ in question 1. With the context removed (or, in this case, added) we here the /k/ as a /g/ because, for all intents and purposes, that context is what’s different between the two sounds.

Question 3

Swap the stops. What do the words sound like now? How do you account for these
transformations? Remember, you did nothing to interchange the nasals in these words.

I suspect that it is now supposed to sound like [stæŋk'gʌmp], but I get something more like [stæŋk'gʌŋmp]. /stæmk/ definitely sounds like [stæŋk] to me, but I can still clearly hear what sounds like an /ŋ/ before the [mp] in `gump’. This clearly, I think, is interference from English phonotactics. There’s nothing /ŋ/-like about this /m/ and there’s nothing /m/ like about this /ŋ/. Incidentally, if I remove one of the pulses from the nasal at the end of `gump’ the /ŋ/ disappears and all I can hear is the illusory /m/. This is incredibly cool.

Question 4

Is it still “damp”?
No, but now it is [stæŋk] (I think that logically follows).

Question 5

When you delete the final burst of “damp” what does the word sound like? Where
did the velar nasal go?

Now it sounds just like /dæmp/ with an unreleased /p/ (I can’t get the upper corner/unreleased diacritic to work). Since there are no longer any signs of the /k^h/ release burst there are is no phonotactic motivation to perceive the /m/ as an engma.

Question 6

Can you get it to sound like “damn”? What does it sound like? Why shouldn’t this
procedure succeed in getting “damn”?

No, I can’t get it to sound like `damn’. Three (related reasons): (1) `damn’ has a longer /m/ consonant in it than `damp’ does. (2) There is too much coarticulatory information from the /p/ closure on this /m/. The way to get rid of it would be to shorten the consonant (see also reason #1). (3) One might expect that we could just chop off the end of the existing /m/, extend the remaining consonant by a dozen or so milliseconds, and have `damn’. Unfortunately, there’s also a lower pitch in the /m/ of `damn’ than in the /m/ of `damp’. I believe this is related to the duration of the segment — in the shorter /m/ there’s just no time to set up a standing wave in the oral cavity so the /m/ in `damp’ is nasal but not as bilabial as the /m/ in `damn’.

Question 7

Can you get it to sound like “gun”? What does it sound like? Why shouldn’t this
procedure succeed in getting “gun”?

No, I can’t get this to sound like “gun” either. I think the explanation is the same: the portion of the vocal tract resonating for the /n/ in `skunk’ is less complex than in the longer /n/ of `gun’.