Fixing Errors

This page (and the associated functionality) is a work in progress, and I have just added it recently (as of 11/23/20).

The /dat/ folder contains a file called formantbounds__template.csv. This file should contain the labels of segments that the user wants to specify hard limits for. Each row contains:

• The label of the vowel.
• Lower and upper limits for F1, F2 and F3.

The functionality is implemented during the get winners stage, that way the analysis never needs to be re-run. When collecting winners, if a file called formantbounds.csv (not formantbounds_template.csv) exists in your /dat/ folder, the function will:

1. Check if the mean F1, F2 or F3 falls outside the boundaries set by the user.
2. If so, the smoothest analysis that does not violate the limits is selected as the new winner. This can potentially be a not very smooth analysis!
3. If no analysis does not fall within the boundaries, the analysis is left as is.

Consider a situation where F2 for /i/ is constantly being misidentified in the 1000-1500 Hz region. This will be immediately visible in the vowel plots created in a folder analysis. The user can specify a lower limit of 2000 Hz for the F2 of /i/ (as in the template) and run the 'get winners' step of the folder analysis again. This will retrieve new winners but exclude all winners with implausible F2 values. This process can potentially fix large numbers of predictable tracking errors quickly.