excellent work. It is a good first step. Please note that you are essentially creating a one dimensional index per person.
Extending that concept leads to categorical truth networks. For instance a network describing, authorities on comedy would not necessarily have any relation to a network describing authorities on physics - yet each person could have their own network definitions uniquely valid to them. A person having a high correlation for two network categories implies a SUBJECTIVE unity of those two categories in the mind of that person. A high correlation between two categories across a group of individuals implies an OBJECTIVE unity of those two categories for that group.
Keep that in the back of your mind as you craft the algorithm for the single dimension case. It is likely that the more general multi-dimension algorithm can be implemented without paying the full cost of the extra dimensions by using clustering similar to sparse matrices.
https://steemit.com/speakfreely/@dsonophorus/i-decide-what-is-or-isn-t-fake-news
Aahhhh, someone who actually understands matrices! Cool! :-)
PS: we are developing UA currently in the slowest possible technical environment - nodeJS - in order to blend-in well with existing APIs. For nodeJS, which I like a lot btw, is not performing "really well" on IO due to its asynchronous and non-blocking nature. But well, if we can make it work here, we can make it work..... (fill out the blanks).