Machine Learning Algorithms Can Predict Emotional Valence Across Ungulate Vocalizations

This repository contains supplemental data associated with the study. It includes the following documents: Data S1 (Audio recordings of calls): A folder containing 3,181 individual audio recordings of animal contact calls. Each file is labeled with metadata including species, context of production, and emotional valence (positive/negative). Data S2: A spreadsheet with detailed information for each call, including animal ID, species, call type, emotional valence, context of production, number of calls, and references to the studies from which the calls were collected. Data S3: A spreadsheet with a summary of the acoustic features extracted before data processing, including metrics for duration, frequency, amplitude modulation, harmonicity, and other features across species, contexts, and valence categories. Data S4: A spreadsheet with a statistical summary of the acoustic features categorized by species and emotional valence, such as the number of observations, mean, standard deviation, median, minimum, maximum, range, skewness, kurtosis, standard error, and interquartile range.