[Expertfinder-dev] Questions on Jei's presentation!

[Li, Jie]

Hi Axel,

Thanks for your comments, which I try to answer below. 

Meanwhile, my thesis has been uploaded under this URL:
http://www.ruleml.org/usecases/foaf/JieLiMCSThesis.pdf
and the ACM classifications can be found under:
http://www.ruleml.org/usecases/foaf/findxprt/acmclassification

The system sources (fact and rule profiles) will be uploaded there:
http://www.ruleml.org/usecases/foaf/findxprt/

Regards,
Jie


> > The SLIDES of Jie's presentation are posted at this URL:
> > http://www.ruleml.org/usecases/foaf/JieLiMCSDefense.ppt
> 
> slide 13:
>   Are these rules meant to be global, or local to ones' scope?
>   What I mean to say is: I'd find it interesting to define such rules
>   local to a context, or linking different RDF+Rules sources, i.e. to
>   define views. However, this does not mean that these rules have global
>   validity, such as rules in ontologies.
> 
>   This is an interesting aspect, since that would mean that rules are
>   COMPLEMENTARY to ontologies rather than concurring.
> 
>   The semantics thant can (similar to SPARQL be defined on top of
>   various possible entailment-regimes, starting from RDF simple
>   entailment on the lower end to more complex entailment regimes taking
>   e.g. RDFS semantics into account, etc.
> 

FindXpRT uses 'global' and 'local' rules, where 'global' here means 'two or more person
parameters' and 'local' here means 'one person parameter': 'global' also means the
system needs to find a connection between the two or more persons globally, across
FOAF-like profile scopes.

Yes, we also use a global ontology, namely the ACM classification (see below).

> slide 14:
> 
>   Where is the evaluation to take place?
>   Imagine I publish such rules mixed with my foaf data on my web server:
>    - Does the client who retrieves my foaf data see the rules or only (a
>      finite subset of?) the materialized data.
> 

We provide a Fact-oriented Normal Form (FNF) and a Rule-oriented Normal Form (RNF) to
our users. Users who are mostly interested in facts (e.g. from the RDF FOAF community)
can choose to compute the FNF, which includes both the elementary facts and all
rule-derivable facts, omitting all rules from the published profiles. Users who are
interested in rules (e.g. for compactness) can choose the RNF, which publishes all the
rules as well as the elementary facts that are needed as premisses of rules, omitting
all derivable facts. Both the FNF and RNF are computed server-side. While the FNF facts
only require look-up, the RNF rules can be executed over their elementary facts either
server-side or client-side.

> slide 22:
> 
>   did you think about which EXISTING vocabularies/categorizations can be 
> reused to describe the areas of expertise in CS or Music? These could be
> relevant for us!
> 

We currently use the ACM classification for describing our CS vocabulary, which has
been
realized in RDFS in my thesis. We don't use a standard taxonomy for Music so far, but
work on a genre hierachy for our www.inDiscover.net could be reused for describing a
categorization of Music.


> slide 25:
> 
>   Where is the information busy derived from? For instance, in our 
> context, this could be done dynamically, by checking the rdf/ical export 
> of peter's google-calendar, etc...
>   We had btw a real demo exploiting such information in our ESWC 
> tutorial on answer set programming rules on top of semantic web data, see
>   http://asptut.gibbi.com/
> 

The information busy is currently stored in persons' FOAF profiles, and we use rules to
extract this information from their profiles. In the future, we should do this
extraction more dynamically, as you suggest.

> slide 27:
> 
> a) What query language do you aim to use? SPARQL is a natural candidate!

We use POSL (http://www.ruleml.org/posl/) to do queries against our rule engine, OO
jDREW (http://www.jdrew.org/oojdrew/), which supports POSL syntax.

> 
> b) Can the same rules also be expressed in e.g. N3, or is your approach 
> in anyway depending on RuleML representation features possibly not 
> available there? (I ask for the easier consumable surface syntax of N3, 
> not knowing too much about the details of N3 rules so far to be honest...)
> 

Rules expressed in OO RuleML or POSL with anchored clauses can be mapped to
N3. For more information, see http://www.ruleml.org/indoo/n3ruleml.html.


> 
> Looking forward to discussions!!!!
> 
> cheers,
> axel
> 
> 
> 
> 
> -- 
> Dr. Axel Polleres
> email: axel at polleres.net  url: http://www.polleres.net/
> 
> 
> _______________________________________________
> expertfinder-dev mailing list
> expertfinder-dev at lists.foaf-project.org
> http://lists.foaf-project.org/listinfo.cgi/expertfinder-dev-foaf-project.org
>