Fathom is a JavaScript framework for extracting meaning from web pages, identifying parts like Previous/Next buttons, address forms, and the main textual content—or classifying a page as a whole. Essentially, it scores DOM nodes and extracts them based on conditions you specify. A Prolog-inspired system of types and annotations expresses dependencies between scoring steps and keeps state under control. It also provides the freedom to extend existing sets of scoring rules without editing them directly, so multiple third-party refinements can be mixed together.

Mozilla’s github

I like the semantic web idea, but it never really picked up. Maybe this will work.