Applause

This repository contains code to run the experiments and to acquire and process the C-SPAN dataset presented in the following paper: Jon Gillick and David Bamman, "Please Clap: Modeling Applause in Campaign Speeches", NAACL 2018

Accessing C-SPAN Data

To download the C-SPAN videos and transcripts, see src/cspan/scripts/download_cspan_data.py. To clean and process the data, see the rest of the files in the src/cspan/scripts/.

Detecting Applause in Audio

See src/Detection/Applause Detection.ipynb. Data to train this model can be found here: https://github.com/hipstas/applause-classifier.

Forced Alignment

We use the forced alignment code built on Kaldi: https://github.com/lowerquality/gentle.

Computing Features

See src/cspan/core. Required libraries for audio features: http://librosa.github.io/librosa/, https://github.com/google/REAPER.

Training Models

See src/cspan/core/Run Models.ipynb and src/cspan/core/Run Neural.ipynb.

Enter Text and Get an Applause Prediction from a Trained Model

Coming soon.