Re:An Airline Geek Trying to Build a Media Giant with No Rep...
by Jeff_Brown on Monday May 28, 2018 @05:05PM
(#17174842)
Google's (now public) PageRank algorithm isn't that complicated, and it brought an astounding improvement in our ability to find what we wanted. There might in fact be a simple algorithm for determining who to trust -- but a closed-source algorithm seems like it cannot by definition go very far to improve media transparency.
If we had more kinds of metadata publicly available -- "[person] trusts [person]", "[person] found [article] helpful for understanding [topic]", "[person] testifies that [person] wrote a good PhD dissertation on [topic]", etc. -- we could publicly experiment with algorithms like, "restrict results to people whose writings have been contradicted fewer times than they have been endorsed", or "to people trusted by people I trust," or "to persons X for whom at least 50 other persons Y testify that X helped them understand [topic]," or whatever.
Hypergraph databases (e.g. Grakn, HypergraphDb) and logic programming languages (e.g. Datalog) already let us express arbitrary metadata -- anything expressible, as far as I am aware.
Getting ordinary users to state their thoughts (and their search queries over those thoughts) in an easily machine-readable manner might be tough. For that purpose I wrote a language[1,2], extremely close to natural language, that allows such machine-readable expression of higher-order relationship information.
If we had more kinds of metadata publicly available -- "[person] trusts [person]", "[person] found [article] helpful for understanding [topic]", "[person] testifies that [person] wrote a good PhD dissertation on [topic]", etc. -- we could publicly experiment with algorithms like, "restrict results to people whose writings have been contradicted fewer times than they have been endorsed", or "to people trusted by people I trust," or "to persons X for whom at least 50 other persons Y testify that X helped them understand [topic]," or whatever.
Hypergraph databases (e.g. Grakn, HypergraphDb) and logic programming languages (e.g. Datalog) already let us express arbitrary metadata -- anything expressible, as far as I am aware.
Getting ordinary users to state their thoughts (and their search queries over those thoughts) in an easily machine-readable manner might be tough. For that purpose I wrote a language[1,2], extremely close to natural language, that allows such machine-readable expression of higher-order relationship information.
[1] Why use Hash: https://github.com/JeffreyBenjaminBrown/digraphs-with-text/b...
[2] How to use Hash: https://github.com/JeffreyBenjaminBrown/digraphs-with-text/b...