What is Praat?
Praat is a software program for "doing phonetics by computer."
Getting Praat
Praat is available for many different platforms (Windows, Macintosh,
Unix, Linux) and can be downloaded (for free) from
http://www.praat.org, and the source
code is also available. There is a Praat users group
http://uk.groups.yahoo.com/group/praat-users.
Intro to Praat handout
This is the
handout (pdf file) for
the Praat tutorial that Kiwako Ito and I ran as part of the LCC
tutorial series in January 2002. NB: The "Control" menu referred to
in the handout is now the "Praat" menu. There is a section on
creating figures by hand, but once you are comfortable with the
basics, it's much easier to use a script to do the drawing for you.
There are many resources for Praat available on the Internet. For
example, many researchers make their Praat scripts available on their
webpages, including:
Mietta Lennes
Hélène Lœvenbruck (in French)
John Tøndering
Kyuchul Yoon
Intro to Praat scripting
Praat includes a scripting language which allows the user to automate
or semi-automate any tasks that can be performed by hand in Praat. A
comprehensive
scripting
tutorial is available at the Praat home page. The same tutorial is
also available from within Praat. In the objects window: Help | Praat
intro. Then Search: scripting.
Scripts
Here are some Praat scripts that perform basic, but important tasks.
You can download (or look at) each script individually, or you can
download a
zip file
containing a set of folders with the scripts described here; the 2002
tutorial; sample sound, TextGrids, image files; some README files,
etc. (N.B.: the very large file
drawing-script-examples.zip
must be downloaded separately. See details below in drawing script
section.) To use the scripts, you will first need to change the paths
(e.g.: E:\PRAAT\TUTORIALS\SOUNDFILES) so they correspond to the
correct locations on your computer.
Labelling scripts
These were written for specific data sets and
semi-automate the labelling process, eliminating (or minimizing) the
need to type labels. They are designed to feed extraction scripts.
Hint: When you run one of the labelling scripts, a little window will
appear in the upper lefthand corner of your screen. Follow the
instructions it gives you (e.g. "Get beginning of target syllable"),
by clicking on the relevant landmark in the waveform or spectrogram,
then click "continue". When you have clicked on all the specified
landmarks, the script will display the waveform, spectrogram, and the
TextGrid it has created. At this point, you can edit the TextGrid to
add labels or correct errors. Then click "continue" to go on to the
next file.
N.B.: If you make a mistake and want to exit the script, click on
"stop" in the little window. But remember if you rerun the script, it
will start at the beginning of the list and overwrite the TextGrids
you have already created. To solve this problem, you can temporarily
put the .wav files which you have already labelled in another
directory. Or you can read in files from a list and edit the list.
label-vowel.praat
creates a TextGrid file with a segment and misc(ellaneous) tier for each
specified .wav file; goes through files, prompting user to click on
vowel beginning and vowel end adds boundaries at specified locations,
displays the labelling to allow the user to make any necessary
corrections (for mis-clicks, for example).
label-tones.praat
(zip) creates a TextGrid file for each specified .wav file;
goes through files, prompting user to click on specified syllable
boundaries. Automatically detects and labels fundamental frequency
peaks and valleys in the target region. The zip file includes sample
files from a study of tonal alignment in Irish by Martha Dalton and
Ailbhe Ní Chasaide, Centre for Language and Communication
Studies, Trinity College Dublin. For more information, see the
project page. See the README file for more details on the script
and examples.
elbow-scripts This set of
scripts (two Praat scripts and an R script), written by
Mary Beckman and
Pauline Welby, implements a procedure for semi-automatically
calculates the position of a fundamental frequency elbow using a
line-fitting procedure. It is appropriate to use such a method when
the beginning of an
F0 rise does not correspond to a
local minimum. See the sound file SET4/elbow-example.wav for an
example of such a case. See the README file for details on using the
set of scripts.
N.B.: The extractF0.praat script was
updated on December 7, 2005 to skip files that do not contain the
specified landmark labels and on February 9, 2006 to specify landmarks
(i and j) corresponding to the example TextGrid file).
For a discussion of the advantages of the technique (at least for some
corpora), see p. 21, Welby, Pauline and Hélène
Lœvenbruck. (to appear).
Anchored down in Anchorage: Syllable structure and segmental anchoring
in French.
Italian Journal of Linguistics. Special issue
on Autosegmental-metrical approaches to intonation in Europe: tonal
targets and anchors, edited by Mariapaola D'Imperio. The relevant
section is also quoted in the README file, on the off chance that you
are not interested in reading the entire paper.
textgrid-creator.praat (zip) This
script was written by
Kyuchul
Yoon. It takes a list of sound files and sentences (orthographic
transcriptions) and creates default TextGrid files (or modifies
existing ones) with these transcriptions. See the README file for
details. This script is very useful for labelling many repetitions of
the same corpus (because it eliminates the need to type the same words
over and over again).
getlabel-scripts.zip
This set of scripts contains four scripts and is almost wholely due to
John
Tøndering.
It also uses a script modified from a script written by
Jackson Liscombe and a
script written by Pauline Welby. Thanks to
Jennifer Venditti for
explaining the workings of these scripts.
This set of scripts allows the user to label TextGrid tiers from a
user-defined menu of choices. It is useful, for example, for doing
complete prosodic transcriptions. This set of scripts should be
particular welcome to those who miss the days of ESPS xwaves
menus. For more details, see the README file.
Extraction scripts
These extract values (durations, formants
values, fundamental frequency, etc.) from files associated with a
TextGrid (label) file. NB: The extraction scripts typically look for
TextGrid files with a certain structure, so a labelling script and an
extraction are often paired.
get-formant.praat
Extracts formant (F1, F2) values from Formant objects (which are made
and removed by this script) at time points indicated in associated
TextGrid files (midpoint of vowel), prints values to a text file
(which can then be read by programs like Excel, for example). (Use with
label-vowel.praat.)
get-values-tones.praat Reads in
sound files, gets time points for syllable landmarks and tones,
calculates durations, creates Pitch objects, extracts fundamental
frequency values and spits them into a text file (which can then be
read by programs like Excel, for example). This type of script is
useful for studying tonal alignment and tonal scaling. (Use with
label-tones.praat.)
Drawing script
draw-waveform-sgram-f0.praat (zip)
This script automates the process of creating figures and saving them
under a variety of formats. Reads in sound files and TextGrid files,
creates related Objects (Spectrogram, Pitch),draws picture in Praat
picture window, saves under specified format (including encapsulated
Postscript and Windows metafile). It includes an option for
overlaying the fundamental frequency curve on the spectrogram, which
is useful for illustrating details of tonal alignment. The black F0
curve stands out against the shades of gray of the spectrogram because
it is framed by a thick white line. The script also allows the user to
control the size of the image. The file
drawing-script-examples.zip
contains two example figures, with the sound files and TextGrid files
used to create them.
Miscellaneous scripts
check.praat pulls up .wav
files and (optionally) TextGrid files; use for checking labels, for
example.
check2.praat similar to
check.praat. check2.praat allows user to check files while masking
file names. This is useful when a blind transcription is required and
the file names could bias the transcriber. The script also can be used
with files that contain decimal points (e.g. dew-2.65.wav) (
problematic for check.praat since decimal points before the one
preceding the file extension are converted to underscores in the Praat
objects window).
save-small-files.praat (zip)
Segments small sound files out of a larger sound file and names them
according to the labels in the associated TextGrid file. Useful, for
example, for creating stimuli for perception experiments.
readin-files.praat
Reads in (opens) all specified .wav files and .TextGrid files. Useful since
Praat does not allow you to read in more than one file at a time.
downsample.praat
Downsamples....
Return to Pauline's homepage
Á venir bientôt, cette page en français !