1. we used the Foo model 2. model: we used the Blah dataset 3. dataset: read our paper on arxiv 4. abstract: data was collected "in-the-wild" 5. buried in paper: we just scraped all the podcasts and videos we could find