The Importance of Language Modeling

So we've seen a lot of interesting things on say unifying structured and unstructured data. So in this kind of thing we're talking about now, I think mostly trying to understand how we go from maybe this kind of idea of observational causal inference or we try to discover causal structure in say an internet scrape of text data. And then there's kind of like, I guess the experimentation side of causal inference. This idea of text is treatment. Have you seen these kinds of studies where they're studying how different ways that you use to present news are treated as treatments?