Shitpost February: eigenvectors

[oulfis]

If I do two of these, does that make it more awkwardly apparent that I'm not actually doing it, or less? Really, I don't think I am ever going to make Dreamwidth my casually-post-stuff place, though I will probably keep trying for the next several years to make it a place I live.

Anyway, I picked the prompt "eigenvectors" because I've actually co-authored research that prominently uses eigenvector centrality, but then I realised that I can't explain what an eigenvector is because I don't really know.

Instead here is some interesting research (in blog form!) from someone who is willing to explain eigenvectors, and even apply them to eighteenth century literature.

Whenever I hear people in the digital humanities beginning to murmur about the hot new thing, word vectors, though, I almost feel despair at how little anybody talks to anybody else -- my mother did her dissertation on word vectors. This is a longstanding field! But that's in computer science, and nobody in literature wants to read a computer science paper.

This post brought to you by: My Career Strategy Of Reading Computer Science Papers Sometimes, To Incredible Acclaim

[inspiration from Shitpost February]

Comments

[jesse_the_k]

All of this sails cleanly above my ability to understand, but I know that
a) eigenvectors is very nice to say (almost as good as rip-rap)

b) that your linked researcher put their work to the test with "Would [software] get into grad school? Evaluating Word2Vec against the Miller Analogies Test” is A1 behavior.

[oulfis]

Yes! My only regret was that he scraped the exam questions from semi-sketchy prep websites, rather than enrolling in the exam itself and either, 1, secretly stealing a copy; or, 2, running the algorithm live on a TI-84 calculator right then and there. It would have been such a wacky nerd heist.

[digsdigsdigs]

I would follow this blog a hundred times over even if your only content was occasionally dragging humanities academics for being incapable of reading outside their disciplines.

[oulfis]

Hahaha, excellent, excellent.

I understand why CS people don't want to read humanities papers -- I think it would be a very good idea if they perused some Ethics 101 textbooks every now and then, but I get that "coding" is the fundamental activity of research for them. But reading things is literally my job!