Comment by throwaway9216 | Hacker News Reader

Comment by throwaway9216 | Hacker News Reader

Due to shitty academic job market, I was forced to take a job part of which could be described as "semantic web evangelist". If you read some incredibly optimistic article about the semantic web and how it solves all life's problems, there's a non-negligible chance that I wrote it.

The semantic web is a colossal waste of time and money. It does more harm than good. Unless you're a savvy PI (probably in a European country) who has used it to lasso a big fat grant from clueless committees who think in terms of buzzwords.

Every successful project which is linked in any way to semantic web, would have been successful without the semantic web.

The only original thing to come out of semantic web, RDF, causes more harm than good. Anyone who has had to deal with encoding sequences in RDF, or any other kind of nontrivial structured data, which is only possible with incredibly asinine "blank node" trickery, either knows this or is deceiving themselves.

Semantic reasoners are either limited to EL ontologies, which defeats the whole purpose of OWL's expressiveness; or else take dozens of gigabytes of RAM to reason over any ontology more complicated than "the pizza ontology". Utterly, thoroughly irrelevant in today's VPS-based world.

The semantic web "community" is about 75% made up of invalid URLs because all these academic geniuses are too busy theorizing about ontologies to realize that all the links into their department webpage are going to stop working when their one-year postdoc expires. Various URL-standardizing services try to address this, unfortunately the number of these services is approximately 1 per semantic web researcher, and in the long run they're no more stable than the department webpages.

The semantic web is a horrific waste of (mostly EU) grant funds that could be used on far more worthy research. The world would be a better place if the whole idea had never been invented.

I strongly disagree. Mostly because the semantic web is very, very useful in the life sciences. And over the next 5 years will save more money being wasted in this field than was ever spend on grants for the semantic web tech. Not just on the academic side but also on the big pharma side.

For the rest I would say this holds true for to much of any CS research e.g.: Grid, Cloud, P2P and GPU compute. Looking at the vast sums spend on this the practical benefits are from outside the academic labs. I.e. all those other buzzwords that are used to justify grants.

OWL reasoning at least standardised the format of many logical constructs and allowed for competitive improvements. Now you can at least change your prover while maintaining your axioms. So you get something like StarDog which is vastly better than the Pellet you are probably familiar with.

I work directly in the life sciences side of it. When the microphone is off, other peoples' interaction with me is basically: "Oh great, we have this data which would be pretty straightforward to store and transmit as JSON, but we have to work with this fucker to turn it into some sort of triple format or something." (Can you tell I don't like my job?)

If you pry beneath the surface the life science ontologies are garbage. Full of subtle errors, disagreement between different ontologies, incorrect changing of namespaces which causes backward incompatibility. You have to have 100% knowledge of the ontology's idiosyncrasies in order to use it with confidence (i.e.: you can't).

Funny you mention StarDog. Semantic web is supposedly this great open data thing that's going to unite the world. Except it's such a miserable community there aren't even any decent non-proprietary tools to satisfactorily fulfill the most basic premises of semantic web, after all these years.

That's just wrong if you're "in the field". Have a look at Fuseki, or HDT, or Virtuoso (hell even Oracle supports RDF). The tools are very mature at this point so I had enough with people saying the Semantic Web never lived up to its expectations because the tools are never mature enough.

StarDog/Pellet are referred to here for their reasoning capabilities, not their "neo4j except more obtuse" capabilities. Fuseki/HDT aren't reasoners, Virtuoso is a partial reasoner but not complete. Don't know about Oracle, they're certainly not non-proprietary.

There are some real and difficult problems the Semantic Web tries to solve, like finding a common language to talk about things and automate information discovery, merging and interpretation.

This may become even more relevant with trends like the IoT (how smart can that be if every company uses its own home-bred standards and the devices can't automatically understand (and react to) each other? This also applies to Robotics / AI.

I'm really not sure the Semantic Web is on a good road there, but they are actually thinking about those hard (and interesting) problems.

There is at least one project where I can see true value, even today: Schema.org (http://schema.org/). And if you're only creating semantic annotations, it really isn't that difficult. Sadly, to smartly combine and make real value out of Semantic Web Data you probably need to be a BIG company like Google.

"Every successful project which is linked in any way to semantic web, would have been successful without the semantic web."

I was thinking many times of it. For most of the time it is true, but not always. I still can find few good examples. Look at dbpedia. It may look messy, but there is vast amount of data there, that is taken advantage of by many projects (also commercial). While i can imagine making a database with the same data, it is the "linked data" initiative that gave it such possibilities.

I agree that there were huge loads of money spent on it, and many projects simply didn't deliver what was promised. But anyway, does research always succeeds in 100% of the cases?

You think if not for semantic web, the dbpedia folks would've just shrugged and said "oh well, if only someone had invented a way for us to share knowledge"?

It's like saying J.K. Rowling couldn't have written Harry Potter if not for Bloomsbury Publishing's in-house style rules.

That's semantic web for you. Dbpedia is a project whose value is 99.9999999999% comprised of its content, 0.00000000001% comprised of semantic web scaffolding, and semantic web comes along and claims full credit for the entire thing.

Yes, but then again, everything would be much harder if there wasn't a framework for that. (As if there wasn't a concept of "publisher" for J.K. Rowling's books. She would need to invent everything from the scratch.

Here's a discussion from a couple hundred days ago that talks about lots of this from some different angles.

https://news.ycombinator.com/item?id=8510401

You should think of the SW more of a thing that exists in nature rather than a piece technology. You seem to use the technology part of the SW and fail to understand that it would exist even if it wasn't thought about.

Reply on news.ycombinator.com