The development of this workshop for audio analysis has been supported by the SpokenWeb (https://spokenweb.ca/) and the Bridging Barriers Good Systems (https://bridgingbarriers.utexas.edu/good-systems/) projects. Developers for the workshop include Brian McFee, Chris Ick, Liz Fischer, and Tanya Clement.

Speaker Identification for Archives. This repository includes several notebooks that walks through the steps of training and running a classifier that takes speaker labels and the audio, extracts features (including vowels), and trains a model and runs it.

This repository contains speaker labels in CSV files for training speaker identification classifiers. These speakers appear in a subset of AAPB files.

This repository contains notebooks with basic and simple workflows for audio processing and analysis in the humanities.

Audio classifier output for identifying Marco Werman as a speaker across all the recordings of The World in AAPBin the American Archive of Public Broadcasting. CSV includes start time, duration, confidence level, speaker name

This repository contains preprocessing instructions for building a universal background model for speaker identification in the AAPB corpus.

A miscellaneous collection of human-approved audio labels.

Data and code for ongoing collaboration between the High-Performance Sound Technologies for Access and Scholarship research group at UT Austin, the WGBH Foundation, and the American Archive of Public Broadcasting.

These demo notebooks demonstrate how to train and run audio classifiers.