Package for testing coherence among listeners tasked with identifying prosodic functions and obtaining cross-correlations between prosodic functions and parameters

The coherence analyses were conduced to verify if these listeners showed agreement greater than chance in their annotations, that is, if the similarities in their annotations are due to the identification of prosodic functions and not to chance. For these analyses, the annotations of each group of listeners were converted into binary tables where lines refer to words in the excerpts and columns to each of the group's five listeners; in these tables, a 1 indicates that a word was identified as a prosodic function by that listener, and a 0 indicates that it was not.

One table was generated for each of the three prosodic functions the listeners were tasked with identifying. These tables are the .txt files in the .zip "coherence_analyses" file. The data for the cross-correlation analyses were obtained through the extraction of values for prosodic-acoustic parameters from the audio files and through the agreement between listeners marking prosodic functions.

The aim of these analyses was to identify the prosodic characteristic of words consistently listened as realizing prosodic functions. The .txt file in the .zip "crosscorrelation_analyses" file is the resulting table, which contains values for prosodic-acoustic parameters for each V-V unit (the unit comprised between the onset of a vowel and the onset of the following vowel) in the audio corpus, as well as the indication if that V-V unit is the stressed V-V unit of a word annotated as realizing a prosodic function by at least four out of five listeners.