I don't see how one can have any hope in a Semantic Web ever succeeding when we haven't even managed to get HTML tags for extremely common Internet things: pricetags, comments, units, avatars, usernames, advertisement and so on. Even things like pagination are generally just a bunch of links, not any kind of semantic thing holding multiple documents together (<link rel> exists, but I haven't seen browsers doing anything with it). Take your average website and look at all the <div>s and <span>s and there is a whole lot more low hanging fruit one could turn semantic, but there seems little interest in even trying to.
I don't think we necessarily need new tags: they narrow down the list of possible into an immutable set and require changing the structure of your already existing content. What exists instead are microformats (http://microformats.org/wiki/microformats2), a bunch of classes you sprinkle in your current HTML to "augment" it.
I include microformats on blog sites, but at scale the challenge with microformats is that most existing tooling doesn't consider class names at all for semantics.
Browsers, for example, completely ignore classes when building the accessibility tree for a web page. Only the HTML structure and a handful of CSS properties have an impact on accessibility.
Class names were always meant as an ease of use feature for styling, overloading them with semantic meaning could break a number of sites built over the last few decades.
There is also RDFa and even more obscure Microdata to augment HTML elements. Google’s schema.org vocabulary originally used these before switching to JSON-LD.
