The Virtual Language Tutor - Teaching Estonian duration

In August 2005 the virtual language tutor was put to use for the first time, at the VISPP summer school, in Palmse, Estonia

In Estonian, a distinction is made between short, long, and over-long vowels and consonants. In addition, these can be freely combined. Out of the nine combinations possible, seven of them are used in the Estonian language.

Perception exercise

ville The sound ‘sate’ can be pronounced in seven different ways to create seven distinct words. In the first exercise the students should learn to perceive the difference between these seven ‘states’. The students could click on the squares in the grid and the VLT would say the correspondig word. The notation used was:
V = short vowel
V: = long vowel
V:: = overlong vowel
C = short consonant
C: = long consonant
C:: = overlong consonant
Once they had familiarized themselves with the map, they could do ‘Exercises’ and ‘Tests’. The VLT would say different words in a randomized order, and the students task was to press the correct corresponding square. Exercise1 and Test1 used the same words as in the map, and Exercise2 used unfamiliar words. During the exercises the VLT told the students if they pressed the correct square or not, and they had to continue until they got it right. On the test they only got one chance to get it right on each utterance, and there was no feedback on whether they pressed the correct square or not.

Vocabulary exercise

ville In the next exercise the students could try to produce the same distinction they had learned to percieve. The vocabulary-practice pane emulates a deck of cards. By clicking on the top card, the student can have the VLT pronounce the word on the card. The student can use the 'Prev' and 'Next' buttons, (or the left and right arrows on the keyboard), to change card. The student should first listen to the word pronounced by the VLT, and then mimic what he heard. The exercise is to try to mimic the duration of the utterance as closely as possible, with special attention on the first vowel and the following consonant of the word. By pressing the record button a recording of the utterance is made, and the recording is analysed. The students recording is segmented into phones, using a technique called 'forced alignment'. Then each phone duration is compared with the VLTs pronunciation. If the duration matches the VLTs pronunciation within a certain threshold, the circle next to the ‘D’ in the top left corner of the pane will turn green. If the duration is above or below the threshold, it will turn red.

Duration analysis

ville In the duration analysis pane, a spectrogram and a time alignment is shown, both of the VLT’s utterance and of the students utterance. The top part of the pane displays the VLT’s recording, with the duration of each phone marked. The middle part is the student recording where there is, in addition to the spectrogram and the phone borders, a pane displaying the offset ratio between the student’s and the VLT’s duration on each particular phone. A rectangle upwards indicates that the student’s duration was longer than the VLT’s on that particular phone, and downwards means shorter.
The exercise is to try to match the VLT’s duration as close as possible. Reducing the threshold setting to the left of the student pane (2) will increase the level of difficulty.
The lower part of the pane offers the possibility to create a modified version of the student recording, where the phones in focus are stretched to match the VLT’s duration(1).
ville Some pictures from the summer school, in Palmse, Estonia