See also: IRC log
scriebick: ericP
scriebic: ericP
<MacTed> scribenic: ericP
<MacTed> scribenick: ericP
<Ashok> scribe: ericP
scrb: ericP (the unix version)
RESOLUTION: approve http://www.w3.org/2010/07/20-rdb2rdf-minutes
<juansequeda> Slides: http://www.slideshare.net/alexmiller/revelytix-rdb2rdf
-> http://www.slideshare.net/alexmiller/revelytix-rdb2rdf Presentation by Alex Miller
alexdeleon: implemented a
federating relational query engine
... worked in XML, XQuery, non-relational systems
<Ashok> Please enter your attendance if
alexdeleon: working in semantic domain at Revelytix
<Ashok> ... I missed you like this: present+:yourname
alexdeleon: working with DOD for cross-domain queries
[slide 3
]
alexdeleon: Business Intellegence/federation are critical
<cygri> strong +1 to everything on slide 3 from D2RQ experience
alexdeleon: given an item in the domain ontology, what db contains it?
<juansequeda> ericP, alexdeleon is not Alex Miller
alexdeleon: if i change X, what will have to change
ericP: when you say "mappings are useful artifacts", are you e.g. sharing the heads of the rules to target federated queries?
alexmiller: not yet, but sounds reasonable
[slide 4]
alexmiller: domain ontologists tend to work on the schema before they have access to the mapping
<Marcelo> Sorry, I was in a meeting.
alexmiller: we built some
closed-world validation tools
... started with SPARQL queries, but that left things we
couldn't check for
... found C&P's talk at SemTech interesting
... we decided to try our own mapping language
... led us to furthur respect d2r
... tx to juansequeda for comparing approaches
[slide 5]
alexmiller: still working on
atomicity of mapping operators
... pattern concept similar to the strings in d2r
... looking for a more flexible way to plug into patterns
... ended up being very verbose
[slide 6]
alexmiller: we view mappings as producing triples
<Marcelo> Where can I find the slides?
alexmiller: each mapping has two
phases:
... .. pull from the data store to produce tuples
<Ashok> Presentation by Alex Miller http://www.slideshare.net/alexmiller/revelytix-rdb2rdf
alexmiller: .. the SPO specifies the shape of the output
<Marcelo> Thanks!
alexmiller: in Revelytix, you can
use one map to refine another
... e.g. base a propertyMap off another propertyMap
[slide 7]
alexmiller: of a SQL query
appear
... all the parts of an SQL query appear
... a propertyMap based on a classMap will inherit the
classMap's subject
... build upon a base uri
... didn't get as far as we hoped. big tension between brevity
and flexibility
[slide 12]
alexmiller: case0 produces
rdf:type triples
... { <http://.../student/#Students<s_id>>
a ex:Student }
Ashok: could you draw on multiple sql tables?
alexmiller: yes, any SQL. also
possible with a property path
... also worked on an ontology to represent a database
... built something that generates an ontology with all the
tables, keys, etc.
... useful to ontologists who can refer to db parts by
URI
... also links database ontology and domain ontology
juansequeda: what exactly generates the ontologies?
alexmiller: we have an oltology
to represent it and some tools to populate it
... uses jdbc, etc.
[slide 13]
alexmiller: propertyMap inherits
from classMap
... { <http://.../student/#Students<s_id>>
foaf:name {.name} }
Ashok: could you embed functions?
alexmiller: yes, need to assign
to a literal and do functions on that
... we defined our SQL, close to '92 or '08.
... functions were problematic; jdbc/odbc standard functions
don't match SQL standard functions
... drivers add their own functions
... we tried to support a superset
... not sure it's appropriate for you, but it worked for
us
... OUTER JOIN syntax was pretty hairy
<Ashok> Eric: Is yr goal to take a SPARQL query and translate to a single SQL query?
alexmiller: we were unable to
push some user-defined functions to the source
... often hard to compute the inverse of functions. e.g.
aritmetic fucntions, to push to client
... we document the capabilities of a source; we support the
complement in the engine
[slide 14]
alexmiller: i liked virtuoso's URI templates based on [db] keys
juansequeda: so in d2rq, you have to write the join, but in your's, you don't?
alexmiller: yeah, that's about
right
... the mapping of foreign keys to primary keys allows us to
compose the joins
... i want to re-use the info in the well-normalized
schemas
ericP: is it acceptable to have a tool interrogate the constraints and generate a configuration?
alexmiller: yes, we provide a default mapping, which you are free[/encouraged] to extend
juansequeda: so an ontologist will be working with the schema, but a user more familiar with SQL would be driving it by join constraints?
alexmiller: pretty much
juansequeda: how much optimization do you do in the query generation vs. leting the SQL engine do it?
alexmiller: ideally, we'd just push it down, but we're all aware that we need to take advantage of the schema and write efficient queries
cygri: per slide 4, you said "SQL
in strings is a bad thing"
... you avoid it in Revelytix by having an ontology which names
the elements in the database? what's the advantave?
alexmiller: easier to validate
using existing SemWeb tools
... but it has its own set of issues; i'm not saying it's the
ultimate solution
<scribe> scribenick: Ashok
Richard: Outlines RDB2RDF approach
Alex: Makes sense ... we want to be mapping language agnostic
Richard: Could we see example of ontology for db schema?
Alex: I will send something
Juan: Marcelo can stert work on a
semantics document
... discussion of form of the document.
... Let's start with a Wiki
This is scribe.perl Revision: 1.135 of Date: 2009/03/02 03:52:20 Check for newer version at http://dev.w3.org/cvsweb/~checkout~/2002/scribe/ Guessing input format: RRSAgent_Text_Format (score 1.00) Succeeded: s/Marcelo/Marcelo and I/ Found ScribeNick: ericP Found Scribe: ericP Inferring ScribeNick: ericP Found ScribeNick: Ashok ScribeNicks: ericP, Ashok Default Present: [IPcaller], Ashok_Malhotra, EricP, MacTed, seema, Souri, cygri, nunolopes, boris, Marcelo_and_I Present: Souri Seema Ashok EricP Juan MacTed cygri Boris LeeF Marcelo Ivan NunoLopes Got date from IRC log name: 27 Jul 2010 Guessing minutes URL: http://www.w3.org/2010/07/27-rdb2rdf-minutes.html People with action items:[End of scribe.perl diagnostic output]