Hoi, While I admire these developments, I find it galling that it is English only. Thanks, Gerard
On 26 May 2015 at 18:03, John Erling Blad jeblad@gmail.com wrote:
It would be interesting to see if some parts of this can be tweaked to allow crowdsourcing of some sort. For example, it is possible to identify some language constructs (patterns) where a better form can be generated. Perhaps an even better SPARQL-form can be generated inline.
Another area where it seems to be possible to improve the results is by adding human expectation. If we ask about "what" we expect a result abut a thing, if we ask about "who" we expect a result about a person, if we use "is" then we want the present value, and so forth. This can be used to filter out spurious values, or even to ask followup questions if an unwanted multiplicity is detected in the result.
On Tue, May 26, 2015 at 5:31 PM, Finn Årup Nielsen fn@imm.dtu.dk wrote:
It seems to be grounded in http://askplatyp.us technology. They give the same answer on the Madonna question.
Without knowing the details I find a text file with mapping here:
https://github.com/ProjetPP/PPP-QuestionParsing-Grammatical/blob/65742217a95...
/Finn
On 05/26/2015 03:58 PM, Markus Krötzsch wrote:
The PPP SPARQL tool is great fun indeed. For example, it does an excellent job at
"Who is married to Madonna?"Quite appropriately, the following question keeps "Loading ..." forever:
"What is the greatest number?"However, I guess it's still a bug ;-)
Will this tool support qualifiers? How?
Also, where does the tool currently take its background knowledge from? For example, how does it know that "married" is a symmetric relation?
Markus
On 26.05.2015 14:13, Ricordisamoa wrote:
Il 26/05/2015 13:40, Markus Krötzsch ha scritto:
On 26.05.2015 02:18, Stas Malyshev wrote:
Hi!
> o If you are not familiar with SPARQL yet, a tool > https://tools.wmflabs.org/ppp-sparql/ has been created by Tpt > https://www.wikidata.org/wiki/User:Tpt and Bene* > https://www.wikidata.org/wiki/User:Bene* to generate SPARQL > queries from natural language questions. > o To do SPARQL queries from the command line, a command line > tool > https://github.com/mariushoch/asparagus was crafted by Marius > https://www.wikidata.org/wiki/User:Hoo_man.
This is awesome! Need more SPARQL tools. I guess the direction to develop this would be to have GUI that allows to build queries (something like we see at https://wdq.wmflabs.org/wdq/) and maybe making ppp-sparql resolved queries - e.g. with actual Q-item for "president" when I ask for "who is the president of the USA?" - which will be
more
efficient.
Should we start some page where we list such tools?
Yes please.
I requested a toolinfo.json to be added to ppp-sparql: https://github.com/ProjetPP/PPP-Wikidata-SPARQL/pull/1 Then you would just search for "sparql" here https://tools.wmflabs.org/hay/directory/#/search/sparql.
I was quite surprised by how well ppp-sparql performs based on very limited information (it's guesses on simple sentences often yield results). Nice work! Maybe it could be more efficient to do some API requests to find the right entities rather than filtering many alternative labels as part of the query. It's not a pattern that we should encourage for production ;-)
Regards,
Markus
Wikidata mailing list Wikidata@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikidata
Wikidata mailing list Wikidata@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikidata
Wikidata mailing list Wikidata@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikidata
-- Finn Årup Nielsen http://people.compute.dtu.dk/faan/
Wikidata mailing list Wikidata@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikidata
Wikidata mailing list Wikidata@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikidata