The analysis of speech data is labor intensive. In this section, the process of transcription and annotation and the software that was used is described.
Two software programs were used:
Both programs are available for free. They are compatible with all operating systems and are easily downloaded. Both provide tutorials. Both programs allow the user to annotate the sound files. This annotation is done through the text tiers, where text can be typed in that is aligned to the speech in the sound file. In the picture below is an example of a sound file opened in Praat, showing a waveform and spectrogram, and a TextGrid that has three tiers in it. Each of the tiers is used for a different kind of transcription. The tiers allow the text (words and letter symbols) to be associated to the sound files. The TextGrids are text files and can be opened in software like Microsoft Word. The user determines the number of tiers they want and what is in them.
This example is a transcription of a North Slavey word bet'ere, spoken by Jane Modeste of Deline, NT, transcribed in Praat using three text tiers. The top tier segments the words into phonemes. The second tier is the word in the IPA, the bottom in the North Slavey orthography. The IPA transcription and the orthography are the same in this example. The tiers in the TextGrid are stamped with the time signatures of the sound files and are saved with the sound files. This type transcription and annotation requires careful work, and it is time consuming. All sound files that were collected will be transcribed this way. We have transcribed about 70% of the sound files that we have collected.
The speech data was recorded in .wav files, and in some cases in MP3 files, though this was avoided as much as possible. MP3 format is not suitable for archiving speech data. Copies of the recordings were given to participants, copies of the recordings were also left in the Language Documentation Center at the University of Alberta. The recordings and annotations and associated metatdata and all resulting databases are archived at the University of Rochester. Currently Dene community members may request copies of this data by contacting us.
The sound files were taken back to the University of Rochester for transcription. This was done from the word lists that were used by speakers in the recordings. In this way, the characteristic properties of the sounds, the way that native speakers pronounce them, can be identified and studied.