Music Supervisor

This repository contains code and data for the experiments in our paper:

Jon Gillick and David Bamman, "Telling Stories with Soundtracks: An Empirical Analysis of Music in Film ", NAACL 2018 Storytelling Workshop. Please cite this paper if you use the code.

To run the experiments yourself, first follow these instructions for downloading additional data.

1. Download the Opensubtitles 2016 data (http://opus.nlpl.eu/OpenSubtitles2016.php) into the root directory of this repo (if you want to put it somewhere else, you'll need to edit load_data.py to change a default path).

2. Use the Spotify API (https://developer.spotify.com/documentation/web-api/reference/tracks/get-audio-features/) to download the audio features for all the track ID's listed in data/spotify_ids.txt. Store the results as .json files in the data/audio_info folder. This folder should have about 96K files with names like 1m4P7RuceKALPVRaX0ZqeN.json.

3. Run parse_subtitles.py to put all the relevant data from OpenSubtitles into one text file for easy loading. This script uses the paths in data/imdb.subtitles.paths, so if you have an error, check to make sure those paths match the path of the OpenSubtitles data on your machine.

4. The run_regression scripts should now be able to load all the data using load_data.py. If you have any errors, check the default paths in the load_data python script.