4 ms·

The semantic web has one huge problem imo: There was/is typically zero incentive to semantically encode data. As shown by things like IBMs Watson, you're better

by lars 14y ago

The semantic web has one huge problem imo: There was/is typically zero incentive to semantically encode data. As shown by things like IBMs Watson, you're better off extracting sentence semantics with NLP than by hoping someone has encoded the data for you. Another thing is that huge ontological databases actually exist, it's just that no one seems to be able to use them for anything useful. Knowledge bases and formal reasoning over them used to be a big deal in AI. This comes from someone who tried to do a startup that essentially tried to incentivize creating semantic data. It's a hard earned lesson, but I now have zero belief that manual formalization of data is ever going to take off. Even if lots of cool things would be possible if it did.

anigbrowl 14y ago
Largely agree. What saddens me is that the extracted semantics are not apparently being shared or exchanged automatically, which is holding us back. On the upside, I remember the view that mass internet usage couldn't take off because open protocols didn't provide sufficient economic incentives for private actors, but that turned out OK. would be quite interested in hearing more about your startup if you want to post to HN or correspond via gmail.
- lars 14y ago
  No prob, I'll send you an email when I get home from work.
mindcrime 14y ago
There was/is typically zero incentive to semantically encode data. As shown by things like IBMs Watson, you're better off extracting sentence semantics with NLP than by hoping someone has encoded the data for you. But yet, adoption of SemWeb tech is growing. Google and Yahoo both rolling out their respective "rich snippets" type interfaces a couple of years ago helped a bit. Those acts raised awareness of the value of SemWeb tech and gave it a bit of a kick. I saw a post[1] earlier today that mentioned some research showing that 25% of the HTML pages out there have embedded structured data (microformats, RDFa). [1]: http://ivan-herman.name/2012/04/18/structured-data-in-html-in-the-mainstream/ http://ivan-herman.name/2012/04/18/structured-data-in-html-i... Edit: better link to specific part of the discussion that mentions the "25% of pages contain RDFa" claim: http://lists.w3.org/Archives/Public/public-vocabs/2012Apr/0018.html http://lists.w3.org/Archives/Public/public-vocabs/2012Apr/00...

by lars 14y ago

anigbrowl 14y ago

Largely agree. What saddens me is that the extracted semantics are not apparently being shared or exchanged automatically, which is holding us back. On the upside, I remember the view that mass internet usage couldn't take off because open protocols didn't provide sufficient economic incentives for private actors, but that turned out OK. would be quite interested in hearing more about your startup if you want to post to HN or correspond via gmail.

lars 14y ago
No prob, I'll send you an email when I get home from work.

mindcrime 14y ago

There was/is typically zero incentive to semantically encode data. As shown by things like IBMs Watson, you're better off extracting sentence semantics with NLP than by hoping someone has encoded the data for you. But yet, adoption of SemWeb tech is growing. Google and Yahoo both rolling out their respective "rich snippets" type interfaces a couple of years ago helped a bit. Those acts raised awareness of the value of SemWeb tech and gave it a bit of a kick. I saw a post[1] earlier today that mentioned some research showing that 25% of the HTML pages out there have embedded structured data (microformats, RDFa). [1]: http://ivan-herman.name/2012/04/18/structured-data-in-html-in-the-mainstream/ http://ivan-herman.name/2012/04/18/structured-data-in-html-i... Edit: better link to specific part of the discussion that mentions the "25% of pages contain RDFa" claim: http://lists.w3.org/Archives/Public/public-vocabs/2012Apr/0018.html http://lists.w3.org/Archives/Public/public-vocabs/2012Apr/00...