Hi,
This an idea of my own. I got to experience Facebook Graph Search since it has
been rolled out and I really like how the idea could work for DBpedia.
What I would like to make is something like Facebook Graph Search, except
simpler and for DBpedia. This might be too challenging to do in the summer and
may overlap with other projects at AKSW, such as AutoSPARQL. But it would be so
awesome. Anyway, as I see it now, the backbone of the realisation would be the
patterns from BOA (by D. Gerber). Some usage examples:
user types in "obama" in the searchwe use the Spotlight lexicalization dataset
or a spotlight interface to find what is the most common sense of "obama" and
return thisuser types in "wife of obama":we use the BOA pattern ?X? wife of ?Y?
to find out we are talking about <dbp:spouse> of ?Y??Y? is "obama", here we can
use the spotlight lexicalization dataset to find the most common sense of
"obama" that fits the type constraints imposed by the BOA pattern usedfinally,
?X? is what we are looking forfrom this, we could construct simple SPARQL
queries to return a list of results
This is the initial idea, it would need refinement to answer real questions
(with "What"s and "Where"s) and queries where BOA patterns can't help
("pictures of obama") and nested queries ("birthplace of wife of obama").
What do you think of this idea? Would it be feasible to do this in the summer?
And what about its politics (autoSPARQL, ...)? And would it be an okay
"product" if it only supported this initial kind of queries powered by BOA?
This idea actually lies pretty close to the DBpedia Better Display idea, maybe
I could try this (in a simple version) if there will be time left (or instead
of another part of the Better Display idea).
Kind regards,
Denis
------------------------------------------------------------------------------
Try New Relic Now & We'll Send You this Cool Shirt
New Relic is the only SaaS-based application performance monitoring service
that delivers powerful full stack analytics. Optimize and monitor your
browser, app, & servers with just a few lines of code. Try New Relic
and get this awesome Nerd Life shirt! http://p.sf.net/sfu/newrelic_d2d_apr
_______________________________________________
Dbpedia-gsoc mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/dbpedia-gsoc
