afcarl / tubetopics

Creates a topic model from a YouTube audio track using Latent Dirichlet Allocation (LDA) and google speech-to-text

tubeTopics

Creates a topic model time course [using Latent Dirichlet Allocation (LDA)] from a YouTube video by downloading the audio transcripts using the Youtube API, or transcribing audio track with Google Speech API (when not available on Youtube).

Note: Requires a Youtube and Google API Key to use.

Download Node and FFMPEG

1. You can download (and install) node here
2. Download FFMPEG (and install)

To set up google APIs

First, set up a Google Cloud account.

Next, create a new project and within the project, enable the Youtube API and the Google Speech API. Create a JSON key for the project and download it.

Then, edit your ~/.bash_profile by adding the following lines:

export GCLOUD_PROJECT=your-project-name
export GOOGLE_APPLICATION_CREDENTIALS=/path/to/your/apikey-XXXXXXXXXXX.json

To use

1. open a new terminal window
2. git clone https://github.com/ContextLab/tubeTopics.git
3. npm install
4. node example/examples.js