<ericP> apologies, i sent my regrets to public-rdb2rdf-wg@gmail.com !

<ericP> i have no idea how my mua elected to default to that domain

Admin Minutes Telephone Telcon on Tuesday? SemTech Panel What is an Ontology? SQL-based approach RDF-based approach Direct mapping Reverse mapping? How to make standard a success

<scribe> AGENDA:

Admin Minutes Telephone Telcon on Tuesday?

SemTech Panel

What is an Ontology?

SQL-based approach

RDF-based approach

Direct mapping

scribe: reverse mapping?
... schema question

How to make standard a success?

SemTech Panel

<juansequeda> What is the usecase for RDB2RDF?

<juansequeda> What are practical experiences?

<juansequeda> What are potential different approaches?

<juansequeda> Why a standard?

<juansequeda> How long is this going to take?

The above are questiine we discussed on the last telcon

<sdas2> Why should I go for it? What are the benefits? Is if feasible? How much effort will it involve on the user side?

Souri: What is it that you cannot do with SQL that you can do with SPARQL?
... integration on data is important ... RDF has no structure
... inference becomes possible
... can create domain ontology with domain knowledge

Ashok: Even a few rules, a liitle inference makes the data very useful

Richard: I can add a view that makes the inference pssible

Dan: What is the source of the semantics?

Juan: Can be a means to an end ... just dreate the RDF

Richard: Three reasons
... Data integration
... Inferencing
... Making data avaialible on the web'
... I can speak about practical experiences with D2RQ

Juan: Triplify has some experience ... e.g. RDF view on OpenStreet map

Souri: Maybe Orri can mention about HCLS

Ashok: Potentially different approaches ...
... SQL approach, D2RQ, ...
... Virtusose RDF views

Souri: People will want to ask about limitations, performance

Ashok: Why a standard?
... Why do we need a standard?

Souri: Data can be fragmented ...
... people will use different databases

Richard: D2RQ works with diferent databases
... Need support tools
... people will write tools around the standard
... ... standard creates a ecosystem

Ashok: How long will it take --- I think between 18 months and 2 years

Souri: How much effort on user's side?

Ashok: There will be tools to help the user

Souri: They will have to buy tools and set them up etc.

Richard: How does this make integration easier than just integrating the data

Souri: database integration is a very hard problem
... we give you a useful approach which helps

Dan: Database integration helps you set the ground for inference
... need metadata

BREAK for 10 minutes

What is an Ontology?

<juansequeda> Slides: http://juansequeda.com/rdb2rdf/Ontologies.pdf

Richard: No ontology case and putative ontology cases are very similar just that there is more information about the URIs that name the classes and properties

Souri: User needs ontology to help write the query

Richard shows demo that shows bits of the RDF graph and help writes the query

Dan: RDF people are comfortable about not being explicit about the metadata

Juan: We already have a glossary, but we all need to be conceptually on the same page.
... we need to explicit about terms and agree

Richard: Be explicit about design space

SQL-based approach

Souri: I will send out the foils to the list
... somewhat modified version of out presentation

Dan: Referential constraints capture domain semantics

Richard: Ontology brings in its own constraints

Souri: we need to speficy Class(Classname, SQLdefString)

Richard: Can I have 2 statements with same classname
... yes, then we will need a union
... Could I have no class?
... You have an OWL had on ... I have RDF hat
... in RDF I can have a node without a type

Souri: Thre types of statements in the mapping language: Class, Property, Constraint
... Discusses subtypes

RESUMING after lunch

<scribe> ACTION: to rcygania2 document issues on the SQL-based approach [recorded in http://www.w3.org/2010/06/20-rdb2rdf-minutes.html#action01]

<trackbot> Sorry, couldn't find user - to

ACTION Document issues on the SQL-based approach on rcygania2 due July 10

<trackbot> Sorry, couldn't find user - Document

<scribe> ACTION: rcygania2 to document issues on the SQL-based approach due July 8 [recorded in http://www.w3.org/2010/06/20-rdb2rdf-minutes.html#action02]

<trackbot> Created ACTION-66 - Document issues on the SQL-based approach due July 8 [on Richard Cyganiak - due 2010-06-27].

<scribe> ACTION: sdas2 to write first draft of SQL-based approach due July 22 [recorded in http://www.w3.org/2010/06/20-rdb2rdf-minutes.html#action03]

<trackbot> Created ACTION-67 - Write first draft of SQL-based approach due July 22 [on Souripriya Das - due 2010-06-27].

<scribe> ACTION: jsequeda to write note on Ways to Leverage Ontologies due July 8 [recorded in http://www.w3.org/2010/06/20-rdb2rdf-minutes.html#action04]

<trackbot> Created ACTION-68 - Write note on Ways to Leverage Ontologies due July 8 [on Juan Sequeda - due 2010-06-27].

RDF-based approach

cygri: I will show examples from code that Eric wrote
... this is from a scalar file that Eric sent around
... uses SPARQL construct quireies from a direct mapping to a domain ontology
... each prefix corersponds to a table
... maps a column to a triple value

Souri: The SPARQL query is written against the direct mapping subgraph
... two steps
... storing this definition and using it to translate the SPARQL query

cygri: Uses SPARQL 1.o so cannot do value translation
... I don't accept that SQL translate is hard to do

Dan: This is for folks who are fluent in SPARQL rathar than SQL

Cygri: 2 downsides, one is the 2 levels of translation. The other is that we need SPARQL 1.1 expressivity
... advantage is that what you get out is very explicit

cygri Perhaps look at an example from SQL-based approach and see how it looks in SQL approach

Souri: Perhaps annotate SQL with what the SPARQL looks like

Ashok: The primary key and foriegn key constructs are not expressed
... can you translate the sql-based example to the rdf-based example automatically

cygri: Should be possible if SQL part is a simple table ... if it is more complex it is not clear

Ashok: To net it out, the RDF-based approach is easier for the RDF/SPARQL expert but ...

1. requires SPARQL 1.1 for expressivity

2. needs two levels of translation

scribe: question about generating the RDF graph from the SQL-based approach

Direct Mapping

Work with Marcelo

<juansequeda> http://juansequeda.com/rdb2rdf/DirectMapping.pdf

Datalog rules to map from SQL schema to ontology and RDF

some database rules sucg as check constraints cannot be expressed

Cygri: Need much less than this for the default mapping in RDF-based approach

- Base table vs. view

Cygri: What can be customized:

- Class name

- Property name

- URI or literal or blank node

- datatype

- langauge tag

c/langauge/language/

Reverse default mapping

cygri: Kevin Wilkinson from HP Labs (jena 2) has a paper to create relational schema from RDF/OWL

Souri: We have a paper on materialized views

What about Object types

Souri: As a first cut not do object types ... just stick to scalar types

Dan: Let's just stick to types that map to XML Schema types

Cygri: Objects can be serialized ... so we can say we just support serialized objects
... need a list of datatypes we support

Ashok: Allow extension points for vendor-specific datatypes

Cygri: Need to be able to handle unknown datatypes
... Need some simple tools to create mapping statements

ADJOURNED

rrsagent make minutes