Overview

Linguistic research sometimes relies on having very accurate formant measurements. Often, our initial attempt at tracking might need modification for several reasons. As a result, the analysis of vowel formant data often runs into the same sorts of problems involving re-doing analysis, data validation, data editing, and so on.

Fast Track is meant to address some common problems that arise when tracking formants (in no particular order):

What analysis parameters should be used for a given sound?

---

Fast track automatically attempts multiple analyses, evaluates them all and returns the winners. This saves the researcher substantial time. Also, all of the alternative analyses are saved for easy corrections.

How can I verify that my analysis is correct? Especially if I need extremely reliable measurements?

---

Images are generated showing the winning analysis, and comparing all candidate analyses, for visual verification.

Where should I measure the formants? Is the measurement reproducible?

---

By default Fast track measures every 2 ms, resulting in a smooth track. Data can be aggregated across sounds in a consistent manner after analysis. Crucially, information about the whole contour is maintained, meaning that the researcher can choose to measure another time point without having to re-analyze anything.

If I think there is an error, how can I fix it?

---

There are multiple ways to fix errors. If another analysis looks better, it can be used instead of the one that was automatically selected. Fast Track also allows for the manual editing of formant tracks.

How can I manually edit otherwise good track with just one bad section?

---

Normally this is quite tedious. However, by relying on Praat's built in formant editors, analyses that require manual editing can be corrected quickly.

Can I modify the behavior of the functions so suit my specific needs?

---

Fast Track is written entirely using Praat scripts with no reliance on any outside software. This makes the modification and extension of the software possible while only being familiar with Praat.

Can the analyses be carried out by people without much training (e.g., undergraduate researchers)?

---

First, relying on Praat makes this software easier to use for most people than anything that relies on, for example, Python or R. Second, the reliance on pre-made images and spreadsheets for error checking makes data validation and error correction easy to do for anyone with the proper phonetic training.