Acoustic Data Pipeline
Convert Praat or VoiceSauce-style TXT files, collapse tokens into normalized time intervals, merge token labels, then clean acoustic CSV files with within-speaker outlier removal and z-score columns.
1. TXT to CSV
Handles tab-delimited text files and removes trailing empty columns.Choose a TXT file to convert.
2. Make Normalized Intervals
Split each token into a custom number of equal-duration bins and average selected measures.
Rows with the same selected token columns are treated as one token.
Numeric acoustic columns are preselected.
The primary output column is measure_mean. Comparison columns can include both versions.
Upload a CSV or use the converted file above.
3. Merge Labels by Token and Time
Attach Full_Label by nearest seg_end within each token, then derive Voicing, Gender, and Context.
Use the shared token/file column, usually Filename or token_id.
Must refer to the same token identity as the data token column.
Gender is the first character of this filename.
Same unit as seg_end. Leave blank to always take the closest label within token.
Upload a data CSV and a label CSV, or use converted/binned data from earlier steps.
4. Clean CSV and Create Within-Speaker z-scores
Select speaker and measure columns, then export a cleaned CSV.
Choose an existing speaker column, or derive speaker IDs from filenames.
Regular expression for derived speaker IDs. Default keeps text before the first underscore.
Numeric columns are preselected. Use Command-click to adjust.
Values farther than this many SDs from each speaker mean.
How removed outliers and invalid zeros are written in clean columns.
Upload a CSV file to configure cleaning.
5. Analysis and Visualization
Generate selectable R code for lmer/emmeans and draw quick mean +/- SE plots in the browser.
Selected effects are joined with * for interactions.
Upload a CSV, or use binned/cleaned data from earlier steps.