JSON-LD and Why I Hate the Semantic Web (2014)

http://manu.sporny.org/2014/json-ld-origins-2/

by walterbell • 11 years ago

58 67 11 years ago

manu.sporny.org

WalterGR3• 11 years ago

It seems silly to see so much stuff being recast as JSON, both formally (standards like this one) and informally.

Growing up in the 80s, I was mostly too young to remember the rise of XML. (Or more: too young to care much about work around a data format.) Presumably a similar thing happened then? I.e. there were existing formal and ad-hoc standards for representing a variety of concepts, and entirely new formal and ad-hoc standards were created to represent the same concepts in XML?

Why not create standards around abstract representations of things, and then create standards for serializing those into concrete formats?! So then we'd already have e.g. Abstract-LD and XML-LD; and JSON-LD would be new. That seems less wasteful than throwing everything out when a different data representation format becomes the new hotness.

(Certainly we learn things about the abstract representations as time goes on. That could be handled by versioning. So maybe there'd be XML-LD 1.0 and 2.0, but JSON-LD would only exist in 2.0 form because nobody was interesting in doing the work to define the JSON serialization of Abstract-LD 1.0...)

jerven4• 11 years ago

Funnily that's what the s semantic web stack achieved. JSON-LD, RDF/XML, turtle, n3 etc etc all different serialisations for the same knowledge. Choose the serialisation on the use case but no need to remodel your data evertime.

haddr5• 11 years ago

Not 100%.. N3 and Turtle don't have the same expressiveness. Then there is XML...

Semantic web is quite easy when it comes to Linked Data (and RDF/S subset), but is becoming painful when more expressiveness is necessary.

Unfortunately, in many cases using only RDF/S is not enough, and some more heavyweight reasoning is required. This is both hard and not scalable for bigger sets of data. Then there is the Open World Assumpltion, which is somehow contrary to what a typical person would expect. You can't easily validate your data for some missing property, because of what OWA assumes: we don't know so we can't tell if it's false...

Anyway Semantic Web has left a nice and mature technology stack, that is great for many tasks. Most of knowledge modelling use cases can be done with RDF and with multiple tools that supports it's creation, editing, reasoning, querying, visualisation, etc.

falcolas• 11 years ago

Yeah, all I could think of while skimming the spec was "Those who do not learn from XML are doomed to repeat it".

Still, this was a great wrap-up discussion, and the specification is easy to read, which is an achievement in its own right and should be recognized.

icebraining6• 11 years ago

Why not create standards around abstract representations of things

That's RDF in a nutshell.

myhnaccount108• 11 years ago

And SGML (which XML is an application of)

nickbauman3• 11 years ago

Hating the Semantic Web is like shooting fish in a barrel ever since Clay Shirky wrote an essay that lays bare the goals as being irrelevant and immaterial.

http://www.shirky.com/writings/herecomeseverybody/semantic_s...

mehh4• 11 years ago

If you agree with the assumptions being made by the author. Even if you do not find it useful in your domain, not sure thats good reason to hate it.

woodman3• 11 years ago

That essay isn't very convincing. All of his examples of logical failures would only work with the most poorly defined ontology and the most brain dead reasoner, which is probably why he described everything using the term "syllogism". For example:

  US citizens are people 
  The First Amendment covers the rights of US citizens 
  Nike is protected by the First Amendment

He then goes on to say that one would conclude that Nike, being a person, has kidneys according to some medical database. First, the ontology: why would you join a medical ontology on "person", instead of something more rational like "human" (you know you're going to have medical data for animals as well)? Second, logical failure: the second statement is not one of exclusivity, you cannot actually link the third statement to the prior two.

I'm sure this guy is very clever, and has plenty of great qualities - but he has built his argument around the failure of logic, when it is clear that he doesn't know very much about it. The use of very poorly defined terms makes it look like he is aware of this...

Retric• 11 years ago

You can't assume automated systems will behave "reasonably".

F=m*a is a great rule most of the time that happens to stop working at various edge cases. Sadly, you can say the same of all of physics and just about every other rule. Even things as 'simple' as living vs dead can get really complicated with lots of odd edge cases.

http://en.m.wikipedia.org/wiki/HeLa new species?

haddr5• 11 years ago

Spot on. That given example neither convinces me. The second sentence is actually false/incomplete and would mislead not only a reasoner, but a human too.

The problem with author's thesis is that he want's semantic web to universally describe the whole world. This is impossible because the world is changing and sometimes contradictory.

OWL for instance doesn't cover such assumptions, and in general most knowledge representation frameworks are monotonic, so it's impossible for instance to withdraw some fact by adding a new one. That what might be a real limitation of the Semantic Web as we know it.

cousin_it• 11 years ago

Another classic essay on this topic is Cory Doctorow's "Metacrap", from 2001: http://www.well.com/~doctorow/metacrap.htm

crusso2• 11 years ago

I hadn't seen that essay before. It seems very dated since the emergence of Google. Many of the problems that he implies are fundamentally insurmountable have been largely addressed by modern search engines.

jplur• 11 years ago

I've been thinking about how to format meta data for my creative files - JSON-LD with my personal @context spec is an interesting idea.

mehh4• 11 years ago

There is more than likely an ontology already defined for the creative meta data, if so you should look at using that rather than inventing your own.

Fannon2• 11 years ago

http://schema.org/ is probably the best choice to go with already established standards. I think there are (even today) some SEO benefits here and for the future maybe some more exciting things.

mehh4• 11 years ago

This was posted before ...

detaro• 11 years ago

https://news.ycombinator.com/newsfaq.html

mehh4• 11 years ago

meh .. yes i know you can repost, but I thought it would be helpful to inform people of this, because they could then see the previous set of comments ... don't know why i bother sometimes.

sktrdie6• 11 years ago

Meh... Semantic Web is a field... it's like hating natural language processing or distributed systems for not having achieved much over the past 10 years. But the fact is they provide strong theoretical foundations for us to develop killer apps on top of.

ma2rten2• 11 years ago

What strong theoretical foundations do they provide? (this is a genuine question)

DyslexicAtheist3• 11 years ago

CSS/HTML, and its syntax validators, DOM, RDF, HTTP, MathML, PNG, heck here is a full list:

http://www.w3.org/TR/

sktrdie6• 11 years ago

Please have a look at last weeks ESWC program (one of the top semantic web conferences): http://2015.eswc-conferences.org/program/accepted-papers

Foundations such as query federation, graph compression (HDT), reasoning, data integration and much much more.

DyslexicAtheist3• 11 years ago

he actually admitted in the comments that the title was clickbait. the value of the article for me was the insight into the inner workings of the W3C.

sktrdie6• 11 years ago

W3C is a standard body. The semantic web is a research field. Different things.

epistasis• 11 years ago

No, it's not, because NLP and distributed systems are producing useful artifacts that help influence the world in positive ways.

Semantic Web is a field that is obsessed with theory, without producing any useful theory, nor any practical artifacts that can be used.

Hating entire fields is an entirely reasonable thing to do when they are not productive and when they draw attention away from more useful activities. And it's quite possible for social dynamics to do that, for the wrong thing to become en vogue, and for really wrong paths to be followed for long periods of time. And it's entirely possible to study something that's actually nothing, such as alchemy. Perfectly fine to hate the field of alchemy.

sktrdie6• 11 years ago

I beg to differ. Being a semantic web researcher myself, I can tell you that it has produced many relevant theories and artifacts that do help influence the world in positive ways. Here's a few to get you started:

http://linkeddatafragments.org/ - allows to query several "semantic web" datasets at runtime by issuing several concurrent HTTP requests

http://www.rdfhdt.org/ - allows to store graph-like data in a very compressed format that is queryable directly in memory

http://events.linkeddata.org/ldow2015/papers/ldow2015_paper_... - (shameless plug) allows you to query and browse semantic datasets (hosted by different people on different servers) on the web using keywords

The fact that they're not useful to you does not make the entire field useless.

bshimmin2• 11 years ago

I'll be honest: "the alter of pedantic technical accuracy" made me snigger a little.

WalterGR3• 11 years ago

Have you read the essay "Metacrap: Putting the torch to seven straw-men of the meta-utopia" by Cory Doctorow? You may enjoy it: http://www.well.com/~doctorow/metacrap.htm

My favorite expression from it: "the lab-coated guardians of epistemology".

More fully: "In meta-utopia, the lab-coated guardians of epistemology sit down and rationally map out a hierarchy of ideas..."

bshimmin2• 11 years ago

I was actually poking fun at the misspelling of "altar" - seemed quite amusing given the context!

jerven4• 11 years ago

Yet in practice schema.org shows that people do put in their webpages and it has a benefit. Cory argues that you can't trust meta data therefore its useless. Newsflash nothing can be trusted, but its at least nice if you can parse it before deciding if its spam or ham!

mark_l_watson2• 11 years ago

JSON-LD is certainly valuable, but I don't understand how to use SPARQL, RDFS or OWL reasoners, etc. with it. Would a client ingest JSON-LD and store it locally in an RDF data store?

There is a ton of great work in knowledge engineering using RDFS and OWL, and tossing that seems like a bad idea.

Rezo• 11 years ago

I think the author's point was that JSON-LD isn't primarily intended for existing users of RDF, SPARQL etc. Most developers already work with JSON on a regular basis (not so much RDF), and JSON-LD is a model for merging and combine data sources, in JSON.

It's good that conversion from and to RDF is now standardized, but it's probably not something that your average dev is going to be taking advantage of. They'll likely continue to ingest, process and store the data in JSON.

mark_l_watson2• 11 years ago

That is a good point.

I am still trying to wrap my head around using JSON-LD. One advantage of RDFS (or OWL) is being able to integrate different data sources using different schemas. That probably is just not a typical JSON-LD use case. BTW, I have written a couple books on the semantic web, so that shows my prejudices :-)

cwyers• 11 years ago

I really wish that, in addition to two videos, he linked some regular text telling us what JSON-LD is and why it matters. Videos, for me, have a much lower information density than text, and are much less amenable to skimming to see if I'm interested enough to go further.

vermooten• 11 years ago

Especially after he slags off so many other writers before him.

vog• 11 years ago

The reply to the very first comment has a link to the JSON-LD spec:

http://www.w3.org/TR/json-ld/

throwaway92164• 11 years ago

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.

haddr5• 11 years ago

"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?

throwaway92164• 11 years ago

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.

jerven4• 11 years ago

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.

throwaway92164• 11 years ago

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.

sktrdie6• 11 years ago

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.

Fannon2• 11 years ago

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.

bane• 11 years ago

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

JSON-LD and Why I Hate the Semantic Web (2014)

Related Stories