On 1/19/13 3:45 PM, John F Sowa wrote:
> On 1/19/2013 2:27 PM, Kingsley Idehen wrote:
>> Why are you comparing SQL (which is used by extensional RDBMS engines)
>> and SPARQL (which is used by intensional deductive DBMS engines) ?
> SQL and SPARQL can be used for either open world or closed world
> databases. It is true that nearly all the stuff on the WWW is open
> world, but many, if not most relational DBs are also open world. (01)
True, but most don't support reference data types, and when they do it
isn't in the form of URIs. Net effect, the keys to the data live in a silo. (02)
Even when they come round to URIs as data types, they will then struggle
(due to mindset) with the notion of decoupling data access protocol from
data representation formats, even worse the notion of an entity
relationship model endowed with explicit machine readable entity
relationship semantics -- where the syntax, notation, and data
serialization formats are all loosely coupled. (03)
>
> In fact, most programmers who have a choice prefer to use SQL notation
> rather than SPARQL notation because is much, much easier and far more
> intuitive to write complex queries with AND, OR, and NOT. (04)
Totally disagree, and from the point of a person that's overseen the
implementation of a DBMS product that supports both. I've lots of live
examples of SPARQL doing stuff that simply can't be achieved "cost
effectively" in SQL -- especially at the scale of the Web. (05)
>
> As for deductions, most of the data on the WWW is so buggy and
> inconsistent that complex deductions are impossible without a huge
> amount of further analysis and clean up. (06)
Yes, and that's where SPARQL shines [1].
>
>> Here is a simple example of a tool showing translation from SPARQL
>> to natural language constructs: http://sparql2nl.aksw.org/demo.
> That tool was *not* designed to translate arbitrary SPARQL queries
> to English. It was developed for the very specific metadata
> conventions used by DBpedia. (07)
No, it will work fine with any RDF based data source. It just so
happened to use DBpedia as an example. I respond from the view point of
a DBpedia project founder. The data is served up using Virtuoso (or
product) and the same would work fine on the LOD cloud instance we host
[2]. (08)
>
> Anybody who designs tools to map NL-like queries to a computable
> form of any kind can and *should* support a translator that uses
> the same metadata to generate an *echo* in the input NL. The echo
> shows exactly how the computer interpreted the NL input. (09)
Fine, where are the SQL bad examples of that?
>
>> I bet you will be very hard pressed to find a similar thing for SQL
>> that's accessible for direct usage on the Web.
> AI and computational linguists have been generating echos from SQL
> since the 1980s. For an example, see pages 3 and 4 of the following:
>
> http://www.jfsowa.com/pubs/futures.pdf (010)
Today, where is a live example of such a thing on the Web? The scale of
the Web is a critical factor in this analysis. Ditto the issues with
data structure, quality, and access protocols etc.. (011)
>
> In the 1980s, they didn't have the WWW. (012)
Yes, so they didn't have the challenges introduced by the WWW. (013)
> But if the DAML people had
> done their duty (i.e., supported SQL as Tim B-L had promised to do
> in the DAML proposal), then queries from controlled NLs could be
> translated seamlessly to and from any mixture of graph or table DBs. (014)
As you know, some of us have implemented products that factor in the above. (015)
>
> That can still be done. (016)
Yes, we've done our part :-) (017)
> But first it's necessary to get the SemWebbers
> to recognize that mainstream IT will continue to ignore them until they
> wake up and pay attention to the requirements for mainstream IT. (018)
Of course, no argument with that reality . (019)
Links: (020)
1. http://bit.ly/UydU9t -- simple example of SPARQL handling disparate
data in a manner that will give SQL fits
2. http://lod2.openlinksw.com -- live instance .
3. http://bit.ly/11ezuCz -- detailed instance report covering loading
and some simple analytics .
4. http://bit.ly/W8MYMj -- Uniprot citations overview .
5. http://bit.ly/13HGvwX -- Bio2RDF entities associated with Genetic .
Dissorders
6. http://bit.ly/101tE55 -- Biopax data sources by collective attributes . (021)
>
> John
>
> _________________________________________________________________
> Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
> Config Subscr: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/
> Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
> Shared Files: http://ontolog.cim3.net/file/
> Community Wiki: http://ontolog.cim3.net/wiki/
> To join: http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J
>
>
> (022)
-- (023)
Regards, (024)
Kingsley Idehen
Founder & CEO
OpenLink Software
Company Web: http://www.openlinksw.com
Personal Weblog: http://www.openlinksw.com/blog/~kidehen
Twitter/Identi.ca handle: @kidehen
Google+ Profile: https://plus.google.com/112399767740508618350/about
LinkedIn Profile: http://www.linkedin.com/in/kidehen (025)
smime.p7s
Description: S/MIME Cryptographic Signature
_________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
Config Subscr: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/
To join: http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J (01)
|