I'm doing my graduate work on that. It's hard. MIT has a cool project looking at how individual terms are covered though: https://www.media.mit.edu/projects/me

I'm doing my graduate work on that. It's hard. MIT has a cool project looking at how individual terms are covered though: https://www.media.mit.edu/projects/media-cloud/overview/ https://www.media.mit.edu/projects/media-cloud/overview/ The relationships when modeling from the author/publication get really big really quick. I suspect as NLP gets better it'll get easier to model though.