4 ms·

Why is it taking so long to invent the semantic web? It could have a large economic impact, but solutions don't catch on. We have a specific way to say "I am a

Why is it taking so long to invent the semantic web? It could have a large economic impact, but solutions don't catch on. We have a specific way to say "I am a teapot", but not a generic way to unambiguously make simple statements like that. At least not a way that has gained traction. What's the missing tech? Maybe there isn't one, and we're just an Eli Whitney coder away from making it happen.

bsder 7y ago
Language is hard. Written language is very hard. Of all the human inventions, written language is one of the very few that wasn't independently invented in many places, multiple times. However, it is so stupidly and obviously useful that anyone who sees it copies it and disseminates it immediately.
- peterburkimsher 7y ago
  I thought written language was invented independently as Chinese, Sumerian, and Mayan. [1] Cuneiform influenced Egyptian hieroglyphics, and the Phoenician -> Latin alphabet. I'll include an honourable mention to Arabic & Hebrew here, which are quite different to Latin, but I think are still related to Sumerian somehow. [1] https://en.wikipedia.org/wiki/History_of_writing#Bronze_Age_writing https://en.wikipedia.org/wiki/History_of_writing#Bronze_Age_...
  - ncmncm 7y ago
    Arabic, Hebrew, and Latin all came literally thousands of years later.
  - azernik 7y ago
    The modern Hebrew (actually Aramaic [1]) alphabet and the Arabic alphabet are both derived from the Phoenician alphabet, which itself was derived (via proto-Sinaitic) from Egyptian hieroglyphs. Egyptian hieroglyphs themselves may or may not be another independent invention; they arose shortly after Sumerian cuneiform, and there was contact between the two societies, but there's no evidence of direct transfer. It's possible that just the concept of writing was transmitted, but no one really knows. [1] The alphabet that archaeologists call the Hebrew or Paleo-Hebrew alphabet is a slight variant of the Phoenician one, just as ancient Hebrew is a dialect on the Canaanite dialect continuum. The alphabet used to write Hebrew since about 300-500 BCE is the Aramaic alphabet, spread throughout the Levant by the Persians when they made Aramaic the lingua franca in their possessions west of the Zagros.
WrtCdEvrydy 7y ago
We have moved from an open web to a closed web so being able to freely move your data around is not a priority to the VC money.
usrusr 7y ago
I'm pretty sure that more 418 have been sent by machines that are not teapots than by teapots. Solving semantics is pointless when truthfulness remains unsolved.
amelius 7y ago
Machine learning fills that gap. There is no need for new formalisms if machines can figure out what was meant anyway.
detaro 7y ago
A bigger issue IMHO is one of motivation. If Google asks to put fragments of semantic data on websites for better search positions, people are more than happy to do it, or at least spend some minimum effort on it. Few are willing to spend effort on it without a clear benefit, but the dreams sold by semantic web advocates require widespread use to be able to build the big cool things. To the contrary, many will see others being able to use data they publish without strict control as a negative. Or their users do: see Facebooks API shutdown. Academic and hobby projects are somewhat of an exception to this, accepting unclear or non-profit benefits, and thus where you find most of this.
nine_k 7y ago
Also, the problem of trust. Lying in an attempt to fool search engines and bring in more traffic is quite widespread on the web as is.
lmm 7y ago
The semantic web is a solution in search of a problem; most websites don't have semantics beyond "the page looks like this", and couldn't be trusted to describe them accurately if they did. Enabling a machine to understand a website in the same way that a human can would be valuable, but is probably an AI-complete problem; in the meantime the best path towards it is probably platforms for easy mostly-automated web scraping or testing (In fact the most useful stepping stone towards a semantic web I've seen is probably adblock, with its simple slider/checkbox interface for telling it "this piece is an ad" in the cases where it failed to figure that out). Maybe "why are Selenium tests so flaky? Can we do better?" would be a good starting point if you want to work on this problem. The inability of most (all?) companies to write good browser-based test suites is effectively those companies' inability to recover the semantics of their own site.
- Tainnor 7y ago
  I never made the link between semantic web and the pain that are browser tests, but now that you say it, it makes sense (especially if you use semantic attributes to identify the page elements for your tests).
miguelrochefort 7y ago
I have the exact same question. I discovered the semantic web almost 10 years ago, and still can't get it out of my head. I keep going back to it. Before writing this post, I was just reading the JSON-LD specification. I spent all weeks reviewing ontologies and wondering why nobody is taking it seriously. I'm convinced that the semantic web (now called "linked data") will be the next big thing. It will happen within the next 5 years. Expect to hear more people talk about it by the end of 2020. Personal health care, quantified self, and IoT communities will adopt it first, followed by AI people. Then it will spread very quickly.
nradov 7y ago
Content creators don't want the Semantic Web because it makes scraping their content easier. For example, there is a standard format for recipes. But most recipe web sites don't use it because they want everyone to read their web site and click on ads. https://schema.org/Recipe https://schema.org/Recipe In order to enable the Semantic Web someone first needs to solve the IP licensing and payment problem. The friction is a huge problem because each little chunk of content has such low value to each consumer.

bsder 7y ago

Language is hard. Written language is very hard. Of all the human inventions, written language is one of the very few that wasn't independently invented in many places, multiple times. However, it is so stupidly and obviously useful that anyone who sees it copies it and disseminates it immediately.

peterburkimsher 7y ago
I thought written language was invented independently as Chinese, Sumerian, and Mayan. [1] Cuneiform influenced Egyptian hieroglyphics, and the Phoenician -> Latin alphabet. I'll include an honourable mention to Arabic & Hebrew here, which are quite different to Latin, but I think are still related to Sumerian somehow. [1] https://en.wikipedia.org/wiki/History_of_writing#Bronze_Age_writing https://en.wikipedia.org/wiki/History_of_writing#Bronze_Age_...
- ncmncm 7y ago
  Arabic, Hebrew, and Latin all came literally thousands of years later.
- azernik 7y ago
  The modern Hebrew (actually Aramaic [1]) alphabet and the Arabic alphabet are both derived from the Phoenician alphabet, which itself was derived (via proto-Sinaitic) from Egyptian hieroglyphs. Egyptian hieroglyphs themselves may or may not be another independent invention; they arose shortly after Sumerian cuneiform, and there was contact between the two societies, but there's no evidence of direct transfer. It's possible that just the concept of writing was transmitted, but no one really knows. [1] The alphabet that archaeologists call the Hebrew or Paleo-Hebrew alphabet is a slight variant of the Phoenician one, just as ancient Hebrew is a dialect on the Canaanite dialect continuum. The alphabet used to write Hebrew since about 300-500 BCE is the Aramaic alphabet, spread throughout the Levant by the Persians when they made Aramaic the lingua franca in their possessions west of the Zagros.

WrtCdEvrydy 7y ago

We have moved from an open web to a closed web so being able to freely move your data around is not a priority to the VC money.

usrusr 7y ago

I'm pretty sure that more 418 have been sent by machines that are not teapots than by teapots. Solving semantics is pointless when truthfulness remains unsolved.

amelius 7y ago

Machine learning fills that gap. There is no need for new formalisms if machines can figure out what was meant anyway.

detaro 7y ago

A bigger issue IMHO is one of motivation. If Google asks to put fragments of semantic data on websites for better search positions, people are more than happy to do it, or at least spend some minimum effort on it. Few are willing to spend effort on it without a clear benefit, but the dreams sold by semantic web advocates require widespread use to be able to build the big cool things. To the contrary, many will see others being able to use data they publish without strict control as a negative. Or their users do: see Facebooks API shutdown. Academic and hobby projects are somewhat of an exception to this, accepting unclear or non-profit benefits, and thus where you find most of this.

nine_k 7y ago

Also, the problem of trust. Lying in an attempt to fool search engines and bring in more traffic is quite widespread on the web as is.

lmm 7y ago

The semantic web is a solution in search of a problem; most websites don't have semantics beyond "the page looks like this", and couldn't be trusted to describe them accurately if they did. Enabling a machine to understand a website in the same way that a human can would be valuable, but is probably an AI-complete problem; in the meantime the best path towards it is probably platforms for easy mostly-automated web scraping or testing (In fact the most useful stepping stone towards a semantic web I've seen is probably adblock, with its simple slider/checkbox interface for telling it "this piece is an ad" in the cases where it failed to figure that out). Maybe "why are Selenium tests so flaky? Can we do better?" would be a good starting point if you want to work on this problem. The inability of most (all?) companies to write good browser-based test suites is effectively those companies' inability to recover the semantics of their own site.

Tainnor 7y ago
I never made the link between semantic web and the pain that are browser tests, but now that you say it, it makes sense (especially if you use semantic attributes to identify the page elements for your tests).

miguelrochefort 7y ago

I have the exact same question. I discovered the semantic web almost 10 years ago, and still can't get it out of my head. I keep going back to it. Before writing this post, I was just reading the JSON-LD specification. I spent all weeks reviewing ontologies and wondering why nobody is taking it seriously. I'm convinced that the semantic web (now called "linked data") will be the next big thing. It will happen within the next 5 years. Expect to hear more people talk about it by the end of 2020. Personal health care, quantified self, and IoT communities will adopt it first, followed by AI people. Then it will spread very quickly.

nradov 7y ago

Content creators don't want the Semantic Web because it makes scraping their content easier. For example, there is a standard format for recipes. But most recipe web sites don't use it because they want everyone to read their web site and click on ads. https://schema.org/Recipe https://schema.org/Recipe In order to enable the Semantic Web someone first needs to solve the IP licensing and payment problem. The friction is a huge problem because each little chunk of content has such low value to each consumer.