Speech file editing with Praat

Goals for this session

In this session you will learn about:

• the phonetics software Praat
• sound objects in Praat, i.e.
• oscillogrammes (or waveforms)
• pitch traces or (fundamental freqencies)
• spectrogrammes
• navigating in audio files
• annotating audio files
• loading, producing and saving Praat sound objects
• creating graphics that you can interpret (and use for your term papers :-))
• inserting those graphics in Word documents

Introduction

Praat is a programme for doing acoustic phonetics with the computer. Praat has been developed by Paul Boersma, a Dutch linguist, and his group. You can find out more about Praat when you visit the Praat webpage.

Another thing that you will need in today's class is SAMPA, a computer readable phonetic alphabet. Really, SAMPA is just another way of writing down the IPA, it basically consists of a mapping of symbols of the International Phonetic Alphabet onto ASCII code. This is convenient because sometimes you may not have IPA fonts available and you still want to do a proper transcript by using your typewriter or in a (text) email.

You can find a list of all the relevant SAMPA characters for English and their IPA counterparts when you click this link. You may want to print a copy of this list, too, it will come in quite handy when you do more transcriptions later in the course.

Exercises

Praat is likely to look different from other computer programmes that you have seen before. Don't get scared though by the unconventional user interface! It is possible to learn how to use it ;-)

1. Download this audio file and save it in your homes directory or on your desktop.
2. Find and open the phonetics software Praat on your computers.
3. Open the audio file in Praat.
   • Click Read - Read from File.
   • Find your audio file.
   • Click Okay.
   • The name of your audio file will appear in the left hand column of the Praat object window.
   • Now you can click Play or Edit and listen to the file.
4. Now mark the beginning and end of each word. This is called annotation. For this you have to
   • Click on the speech file,
   • then on Label&Segment, select to TextGrid,
   • remove the john, mary, bell and call your tier words,
   • click ok.
   
   You have now created a TextGrid. Now select both the speech file and the textgrid in the Praat objects list and click Edit. For the annotation,
   • Find the beginning of the first word.
   • Click on it in the spectrogram. A red line will appear and a circle in the TextGrid.
   • Click into the circle. You have created the first boundary.
   • Put boundaries at the beginning or end of each speech sound.
   • You can delete (Alt+BackSpace) and move all boundaries (click and drag).
   • In between two boundaries you can write the word.
   
   You can save your work (only the TextGrid) by clicking on it in the Praat objects list and saying "Write to file".
5. When you finished with the orthographic annotation of the words, add another tier below the words tier. You do that by clicking Tier and Add interval tier in the TextGrid Window. Name your tier IPA. Do the same thing over again and name the next tier SAMPA.
6. Transcribe your words in IPA and in SAMPA. You will find a way of inserting IPA symbols when you click Help - Phonetic symbol chart in the TextGrid window. The ASCII-like IPA symbols can be inserted like that, some other symbols have to be sort of described, e.g. for the voiced dental fricative you have to write \dh and the appropriate symbol will appear.
7. Now try and generate the pitch track (something like the melody of the sentence)from the audio file. To do that, go back to the Praat object window and click on the audio file. Then click Periodicity - To Pitch. Listen to the pitch! Do you recognise the melody of the sentence?
8. Draw the pitch graph! To do that, click on the pitch object, go to the Praat picture window, mark an area and then click Draw in the object window. Now put the textgrid on top of that in the same way. Make sure that you can see both the pitch and the textgrid well. When you finished, save your graphic as a Windows meta file (click File - Write to Windows meta file). You can insert a file like that into any word document simply by clicking Insert, Graphic, From File and then doubleclick on the file.
9. Now extract a single word from your sentence (please choose a "medium-size" word like heaven or nothing). Mark a selection (e.g. by doubleclicking between two boundaries or by drag and drop), then click File, Extract sound selection. Your selection will appear in the objects window. You can rename it so that you will find it again later. Create a new textgrid (cf. above) and now mark the boundaries of the individual speech sounds and annotate the speech sounds.
10. Generate a spectrogramme from your new audiofile with the word in it. Like with the pitch, draw or paint the spectrogramme and put the textgrid on top of it. Insert this into a word document, too.

This is a lot of input for one session and you will have the chance to practice working with Praat in the tutorial. Also, you will find Praat tutorials on the web which you may find helpful.

For reference: Finished TextGrids

Click here for a complete orthographic, IPA and SAMPA annotation of the phrase you annotated in class.

Homework

For next week, please read Clark and Yallop, sections 7.1, pp 207-209 (or in the old Clark and Yallop sections 7.1-7.4, pp. 183-194).
Also, familiarise yourselves with the following concepts so that next week you will have an idea of what we'll be talking about:

• vibration or sound
• wave and sine wave
• amplitude
• frequency