From: ddurian@ling.ohio-state.edu
Subject: Week 2 Summary/Week 3 Planning
Date: Mon, 07 Aug 2006 16:36:40 -0400

August 2nd Meeting

The principle task that we accomplished during our first meeting
session was determing the PRAAT scripting needs for my study of vowel
variation among White middle and working class speakers in the Columbus
area. Previously, I had begun taking F1, F2, and F3 measurements from
tokens containing the vowels /ow/ and /aw,/ two dipthongs undergoing
interesting sociolinguistic variation in Columbus. These measurements
have been taken by hand from 20 tokens by either me or one of the two
interns who have worked on my project during the last year.

The scripting needs for my project have evolved from this previous
work, in that I'm basically looking for a way to automate parts of the
token location and F measurement extraction routines. Hopefully, the
scripts will also help me and my interns more accurately measure vowel
variation, as automatic extraction would eliminate bias due to
interrater reliability issues.

The script we developed allows me to a) create three tiers that are
highly useful for the token location and measurement task. The first
tear is used for sentence by sentence transcription, so that the word
token chosen can be seen within it's sentential context. The second
tier allows the word token itself to be marked, so that surrounding
phonetic contexts in which a given vowel occurs can be noted. The third
tear allows the vowel itself to be marked. This is useful because it
provides a target for a second script routine to go throw and take
F0-F3 measurements for the vowel.

After we developed the script, we tested it on data obtained from one
female speaker from my extent 32 speaker corpus. The vocalic variable
targeted in the tests is /uw/.

Following the test of the script, Mary and I discussed normalization
studies that have been completed in phonetics over the last 15-20
years. This discussion was prompted by my consideration of F0
extraction from my vowel tokens. F0 extraction is being considered
because many normalization algorithms use F0. Normalization is
important for my project, because I will be making comparisons among
differnent subgroupings of speakers--by sex, and within sex, by age.

At the completion of the meeting, I went away with several studies 
to look at concerning normalization. These studies are the following:

* Johnson, Keith. 1990. Contrast and normalization in vowel perception.
Journal of Phonetics, 18:229-254.
* Johnson, Keith. 1990. The role of perceived speaker identification in
F0 normalization of vowels. Journal of the Acoustic Society of America,
88.2:642-654.
* M?nard, Lucie, Jean-Luc Schwartz, and Louis-Jean Bo?. 2004.Role of
vocal tract morphology in speech development: Perceptual targets and
sensorimotor maps for synthesized French vowels from birth to
adulthood. Journal of Speech, Language, and Hearing Research,
47.5:1059-1080.

August 9th Meeting: Planning

The principle tasks I hope to complete in the August 9th meeting are:

1) Developing a PRAAT script that will extract F0, F1, F2, and F3
measures from the vowel tokens that are located using the script we
developed at the August 2nd meeting.

2) Discussing the results of the 3 normalization studies I read after
last meeting, as well as their relevance to my project.