I recently saw a demo of Invention Machine, which does semantic extraction of a body of data. They integrate it with engineering processing tools, and their tar

I recently saw a demo of Invention Machine, which does semantic extraction of a body of data. They integrate it with engineering processing tools, and their target market is companies that have a process that they can improve. The problem with semantic search is that out of context, it doesn't buy you much. Note how many of the demos of Powerset have been on a tiny body of data. I'm very familiar with the state of art in natural language processing. Invention Machine is pretty good. Powerset seems pretty good. It is silly to think that Google, with a mass of excellent engineers and folks like Norvig heading the effort, will be beaten because Powerset will go public with semantic search features sooner. Google certainly has people working on it. Even a marginal improvement on search or adWords relevance would get them millions. Unless they are already a sloth of a big company, there is a reason such features aren't released. My guess is that the techniques aren't good enough yet to take on the entire web.