We wanted to take an arbitrary graph of links on the web and take it though a network reduction that yields an arbitrary graph of sets of keywords. From there we can do key word subselection and generate the user search semantics. The syntax for all this was user search form is simply a partially ordered list of key words. The search engine was to create a canonical semantic graph that covers the domain of web key word links with user semantics.
In this scenario we have the broswer and the web industry. The web industry is Wiki, Google, Yahoo, facebook, MySpace, Google+, Groupon, Amazon,....) That industry produces the range of key word sets, and their links, that is, all possible paths through the web, traversed by key word. Then the browser runs a javascript personalized, maximum entropy, key word encoder; governed by the interaction between user search requests and the browser personalized Imagisoft Search Gadget!.
We get a surfboard to ride the webosphere. I hereby patent the idea. This is nothing more than the DOM tree analyzers, Jquery, simple graph traversal stuff, stuff I do on this blog now and then.
For example. The Search Gadget automatically learns that the appearance of Wiki in the search query implies a retrieval of a Wiki subgraph. The search gadget is doing entropy analysis pon user key word lists, discovering random appearances of universal keywords. It is discovering groups of key words the user frequents. The user and the gadget together work toward a maximum encoded search forms, little canonical graph segments of sub selected key words. These subsegments that can be pieced together or parallel encoded for actual retrievals. We get bits of search grammar, a boolean logic os some complexity that the user and gadget settle on.
No comments:
Post a Comment