I think part of the problem is defining what the "Semantic Web" or "Semantic Publishing" is. For me, it is being able to navigate information based on semantic content. For example, applied to web search, I'd expect the search engine to be able to present me with the topics present in my search results and allow me to re-rank/refine those results based on the presence of topics. If I search for cancer, I would expect the search engine to identify the topics within my search results (lets say: diagnosis, treatment, survival, research etc.) and let me rank results based on the relevance to topics. If I was interested in research, for example, then I'd indicate that via the interface and the results would update. This would seem obvious when it's a search for a term that you already know something about, like cancer, but it becomes powerful when your searching for something you don't know anything about. It allows you to learn and navigate much quicker. It's also easy to see its value when you extract this out to being able to navigate your social media and news stories.

There are startups today that have the technology to do this. Some rely on machine learning (including deep learning) to pre-build models that are then used to classify text based. Others build topic/concept models "realtime" based on the content you pass it. And others still rely on a linguistics approach (admittedly, not many recent companies take this approach for various reasons). One even has a tool that will let you search Google in a somewhat similar fashion to what I described. However, most are still looking for funding and some have been pulled away to focus on customer feedback/BI rather then the semantic web. Also, it can be difficult to start such a company given the amount of resources it needs to get going. Very difficult to bootstrap.