Speaker unverified: belief

Transcript context

…Can you describe what kind of features can be extracted in this way? Well, so it depends on the model that you’re studying. So the larger the model, the more sophisticated they’re going to be. And we’ll probably talk about follow up work in a minute. But in these one layer models, so some very common things I think were languages, both programming languages and natural languages. There were a lot of features that were specific words in specific contexts, so the. And I think really the way to think about this is that the is likely about to be followed by a noun. So you could think of this as the feature, but you could also think of this as protecting a specific noun feature. And there would be these features that would fire for the in the context of say, a legal document or a mathematical document or something like this. And so maybe in the context of math, you’re like the, and then predict vector or matrix, all these mathematical words, whereas in other contexts you would predict other things, that was common. And basically we need clever humans to assign labels to what we’re seeing.…

Stored transcript either side of the excerpt. The highlighted words are the published quote; the surrounding text is unedited source, never generated.