Preparing the data
Perhaps the easiest way to understand the file format that R requires is to look at the sample data file.
The data file has columns, with header rows word, token, Time, F1, F2, and F3.
The word column indicates the name of the item (which is not necessarily a "word").
The token column indicates the token number (each item should have multiple tokens).
Note that there are multiple entries for a word and token combination. For instance, token "1” of the word "bab” has 31 rows. These rows contain the formant information over time. In this case the times range from -0.225 to 0; at each time point, there are three formant values. This specifies the formant values for the duration of the vowel. (You plot the formant values against Time to check this, using Excel or something similar).
Things that can vary
- The time values do not need end at zero; they could start at zero; or range between two other values altogether.
- The formants don't need to be sampled at regular intervals.
- It's possible to omit formant values (for instance, F3) from the data file. But see the special instructions for this in making comparisons.
- The formant values can be in any quantity (Hertz, Mel, Bark...)
- The time values can be in any quantity (seconds, milliseconds...)
Things that have to be the same
- The header row needs to be exactly the same as in sampledata.txt (excepting that you can omit a formant)
- The time markers between vowels need to be commensurate between tokens. The time can represent seconds from the start of the vowel, or seconds from the end of a vowel. But it should not be the times of the vowels in the full audio file, for instance (so that one vowel would start at 100 s and another at 200 s).