I'm not sure if Graphene in specific could do it as it is now but I do think it could be upgraded to manage it. I know for sure Tauchain could do it which is why I've given so much of my attention to Tauchain. Freenet is a bit outdated (2001) technology and doesn't even handle basics of the semantic web so I would not rely on it. The technology you need is the technology behind the semantic web which Tauchain promises to have. Solid also looks good and is made by Tim Berners-Lee, designed specifically for social media.
Generally though, it's going to happen sooner or later whether we see it on Tauchain or Graphene. To me a blockchain is just a data structure. But it does not have to be rigid, you can make it flexible and Tauchain shows that, but even to some degree you see it in DPOS. The blockchain allows you to have a flow of time, a sequence, an agreed upon shared state for any specific point in time, and for certain things this is necessary, but you can use other data structures too like tangles(DAGs) or even what SAFE Network is doing, depending on what you need to do.
Steemit needs the data structure it has but you can do sidechains with it to extend it.
First off, I completely respect your opinion and I want you to know that I am really loving this conversation. These types of debates are why things like steemit were invented. One day someone is going to look back at this discussion in the same way we search usenet archives and go "my god if only they knew!" :)
Don't knock freenet, just because it's older. None of us are spring chickens anymore and the next generation is looking as abysmally hopeless to us as we did to our parents :D
It's been around for awhile yes, but it does receive a lot of continuous development efforts. And besides, what I was trying to say is you use a system like it for proper raw data storage and then keep your contexts and reference sets in a db of some sort, blockchain or otherwise calling into freenet when you need to retrain or need to pull out something larger that wouldn't be adequately stored on the blockchain.
That said, I really don't think that a semantic web is the correct answer here or even on the right track.
For example, we've had HTML5 for years now and it already has a TON of semantic features, but they are hardly used at all. To this day, we still swim in div soup. I just interviewed a web developer who had never heard of article or aside tags or what their intent is. Despite the HTML5 certified expert on his resume.
But that's not really my point.
The problem is it's all but impossible to do semantic analysis when nearly all commentary is moving to markdown style syntax, which is reverting right back to the bad old days of HTML where we mixed style with structure.
What you need to do if you really want to move in this direction is begin with parsers and classifiers and build more freeform knowledge extraction systems.
The Apache cTakes project is a perfect example of what I'm talking about here. Take a Doctor's free form notes live and in real time and turn it into a properly coded PMR.
http://ctakes.apache.org/
The job of the experts in a system like that is to train the classifiers until they are "good enough" to handle the load. Devolve the task into more and more specific subclassifiers maybe using a genetic algorithm until you have a complete ontology similar to Cyc.
http://www.cyc.com/platform/opencyc/
You don't need hard core experts to such a system, just a bunch of people who can say yes, no or uncertain on anything the RNN comes up with as it's training. So you have the net itself and then you have "what the net knows", very similar to our own brains where we have regions that code and store information and other regions that act on information.
Come to think of it, Cyc would make a good candidate for your Alexa replacement. If we could find a way to build even something like that, organically in an automated, if assisted fashion and find a proper method of distributed storage.
I can see how the idea would work powerfully.
You mention a dag, and I mention node4j which is a specific graph DB, so I think at the most important levels we are actually talking about the same thing but from different angles.