I meanwhile found a public accessible link to your publication:
http://www.mpi-inf.mpg.de/yago-naga/yago/publications/aij.pdf in which you
write:
"However, in contrast to the original YAGO, the methodology for building YAGO2
(and also maintaining it) is systematically designed top-down with the goal of
integrating entity-relationship-oriented facts with the spatial and tempo- ral
dimensions. To this end, we have developed an extensible approach to fact
extraction from Wikipedia and other sources, and we have tapped on specific
inputs that contribute to the goal of enhancing facts with spatio-temporal
scope. Moreover, we have developed a new representation model, coined SPOTL
tuples (SPO + Time + Location), which can co-exist with SPO triples, but
provide a much more convenient way of browsing and querying the YAGO2 knowl-
edge base. " etc. page 3
so it seems one special feature is the explicit treatment of space and time,
which sounds interesting. So I would like make some of my questions more precise
"YAGO has about 100 manually defined relations, such as wasBornOnDate,
locatedIn and hasPopulation. Categories and infoboxes can be exploited to
deliver instances of these relations. (p.3) "
...
"The new YAGO2 architecture is based on declarative rules that are stored in
text files."
-Is there a wiki or some other public accessible place for those like the
mapping wiki of dbpedia?
-what do you do if infoboxes change?
-How do you treat microformats?
"Instead of seeing only SPO triples and thus having to perform an explicit
de-reification join for associated meta-facts, the user should see extended
5-tuples where each fact already includes its associated temporal and spatial
information. We refer to this view of the data as the SPOTL view: SPO triples
augmented by Time and Location. We also discuss a further optional extension
into SPOTLX 6-tuples where the last component offers keywords or key phrases
from the conteXt of sources where the original SPO fact occurs.(p. 20)"
It is not yet fully clear to me how your concept go together with other
concepts to include contexts like with named graphs or with the inclusion of
context-ontologies within formats like JSON-LD, eventually that would need a
longer discussion, are you planning to set up a wiki page on datawiki, like for
example there is one for JSON-LD:
http://meta.wikimedia.org/wiki/Talk:Wikidata/Data_model/JSON ?
Is it possible to extract the Yago queries in some RDF serialization format?
_______________________________________________
Wikidata-l mailing list
[email protected]
https://lists.wikimedia.org/mailman/listinfo/wikidata-l
