I’m now at the Max Planck institute for Human Development, Berlin.
I’ll be working on the large Knowledge Collider project (), an EU grant that aims to create a platform for real-time reasoning on the semantic web.
The semantic web is being around for years, but only now it’s getting enough momentum. This is an area where I think what we know about semantics in psychology may make an important contribution.
This is called web 3.0 and is supposed to be the next big thing.
What’s interesting about Web 3.0
The basic idea is that the web is written for humans who read text; but it could be a lot more useful if it was also written for machines that can ‘reason’ on structured information. Right now, a machine can do much inference on propositions. The basic unit on the semantic web is the RDF triplet, which is basically a [SUBJ, VERB, OBJ] proposition.
Many existing databases are converted to a format that is, basically, a propositional analysis. But most of the web is, of course, plain text and not amenable to straight conversion between some other structured format to RDF. There are ongoing efforts to do this automatically from plain text. And, as happens with ‘automatic propositional analysis’, it’s not very good. But the fact that some companies do have databases (which are structured information, and easily translated into RDF triplets) is very promising.
Imagine that a spider for a search engine could read on the amazon webpage:
[JoseQuesada, likes, Kintsch-12345]
[Kintsch-12345, hasTitle, Comprehension]
Or from the wikipedia page:
[Kennedy, said, Prop-9876]
Prop9876: [Ich, bin, ein-Berliner]
Then, a machine could draw a lot of interesting inferences; and browsing the web will not need to be reduced to keyword matching: people could construct queries in a language that operates with RDF (SPARQL) or we could make search engines to translate natural language into SPARQL.
If all this sounds to you like ‘making the web a humongous database’ or ‘the return of good old-fashioned AI’, you may be right.
The challenge now is to write something that can do reasoning with trillions of propositions in something close to real time. This larKC project puts together people from parallel computing, cogSci, and hardcore reasoning people to create a service like that. It has funding for 3 years.
I will be posting more on semantic web stuff here (when appropriate) but mostly on the .