The basis for the vocab ordering project was the realization that even more valuable than learning a common word is learning a common word that shares a lot of sentences with a lot of other common words.
Imagine a network of "knowledge states" where arcs represent what you need to learn in order to move from one level of knowledge to another. It was the idea of finding a path through such a network that eventually led me to make the connection with the traveling salesman problem.
It just occurred to me this evening that there may be a connection with Google's Page Rank algorithm too. After all, it's all about finding the links to the pages judged most valuable because they have links to the most valuable pages and so on.
As the wikipedia article says: "the algorithm may be applied to any collection of entities with reciprocal quotations and references".
Incidentally, I was at the conference where the PageRank paper was first presented. It was WWW7 in 1998. I ran the XML tutorial and sat on a panel on hypertext. I can't remember if I attended the Brin/Page talk or not, though. I do remember hanging out with Ted Nelson (and having dinner with him and Roger Clarke) and meeting Rohit Khare and Adam Rifkin for the first time there.
The original post was in the categories: algorithms conferences google speaking but I'm still in the process of migrating categories over.