First, the Praat manual is a useful resource for figuring out how to use Praat. The manual is available online and offline as part of the Praat software. You can access the manual in Praat by going to Help > Praat Intro. If you are looking for something specific, you can use Help > Search Praat manual...
Speech Annotation
- Systematically divide speech material into parts
- Facilitates analysis
- Easier to locate and compare tokens
- But the process can be time-consuming and boring
- Some common kinds of annotation: sentence/word boundaries, phoneme boundaries, parts of each phoneme, acoustic measurements
Praat Basics
When you open Praat, two windows pop up. One is called Praat Objects, and the other is called Praat Picture. In most cases you can ignore Praat Picture and close it.
In the Objects window, there is a fixed menu on the top (New, Open, Save). This menu does not change. On the left side is the list of objects. On the right is the dynamic menu. This menu changes depending on which objects are selected. On the bottom, there is a list of fixed buttons.
Types of Praat Objects
- Sound
- A short sound file. Typically words and utterances. You can read in sound files through Open > Read from file...
- LongSound
- What it sounds like. A sound file that is up to 2GB long.
- TextGrid
- An object used for annotation of the sound. You can create TextGrids from both Sound objects and LongSound objects.
There are also other types of objects, but you don't need to worry about them when starting out with Praat.
The dynamic menu on the right side of the Objects window will change depending on what type(s) of objects you have selected.
Deciding on a set of symbols to use
You can use IPA symbols for annotation, but this is often more time consuming and you may run into character encoding issues. Often, phoneticians will use a set of symbols that contain only ASCII characters. Arpabet is such an example for English. For example, the CMU Pronouncing Dictionary uses a version of symbols based on Arpabet. If you decide to not use the IPA, you should look to see if there's a standard ASCII transcription system of the sounds in your language. If not, you can create your own ASCII transcription system, but be sure to keep track of which characters stand for which sounds.
Creating a TextGrid
To create a TextGrid, first select the Sound or LongSound you want to annotate, then click Annotate - To TextGrid...
A window will come up to ask you for the tier names and which tiers are point tiers. What tiers and how many tiers you need depends on what you want to study, but it is safe to start with just "word" and "phoneme". Interval tiers are useful when you want to know where and how long a segment is. Point tiers are useful for recording measurements/information taken at a specific time.
Now that you've created a TextGrid, select both the Sound and the TextGrid, and click View & Edit. This will open both objects in the same window.
The window looks something like this:
Annotating Speech Data
Now that you have a TextGrid, you can start segmenting the file by adding boundaries to the interval and point tiers.
To start segmentation, first zoom in to the part of the sound file you want to segment. You can do this by selecting a part of the sound file, and then clicking the "Sel" button on the lower left corner (or use the shortcut Ctrl+N/Command+N).
First, decide where the boundary should be. You can do this by listening to the sound and consulting the waveform or spectrogram. To add a boundary, simply click the small circles on the tier where you want the boundary.
Hold down your mouse key to move a boundary. You can move aligned boundaries on different tiers by also holding down the Shift key.
To delete a boundary, select the boundary you want to delete and go to Boundary > Remove (or Alt+Backspace).
When studying vowels, one thing you might want to do is to label the F1 and F2 values. You can do this by making sure that "Show formants" is checked:
The red dots are formant tracks. You can get the formant of a vowel at a certain point by going to Formant > Formant listing. Now you can add the formant values to the F1 and F2 tiers.
Save frequently! This can be done in the TextGrid Editor window. Just do Ctrl+S or Command+S, and this will bring up a dialogue that allows you to specify the location where you want to save the TextGrid. Praat sometimes crashes, and you don't want to lose your hard work.
Running Praat Scripts
To open a Praat Script, go to the top menu and select Praat > Open Praat script...
This will open the script in a Script window.
To run the script, click Run > Run. Or use the shortcut Ctrl+R/Command+R.
Useful Praat Shortcuts
| Windows | Mac | Function |
|---|---|---|
| Tab | Tab | Play from cursor/Play selected portion |
| Ctrl+I | Command+I | Zoom in |
| Ctrl+O | Command+O | Zoom out |
| Ctrl+N | Command+N | Zoom to selected portion |
| Ctrl+A | Command+A | Show all |
| Ctrl+B | Command+B | Zoom back |
Simple Manipulations of Sound Files in Praat
You can edit sounds in Praat pretty much the same way you edit text. After selecting a portion of the sound, you can cut, copy, and paste as you would text. In the Edit menu of Praat, there is also the option of Set selection to zero, which set the elected portion to silence. You can also reverse the selected sound with Reverse selection.
Save as binary file
If you're working with multiple objects in Praat, you can choose to save all of them together as a binary file. The binary file has the extension .Collection. The next time you read this file into Praat, you will see all the objects as they were when you saved them.
