[PragmaticWeb] Feedback

Sebastian Samaruga ssamarug at gmail.com
Tue Nov 22 18:39:39 CET 2016


William, I'll respond inline for answering your questions:

On Nov 22, 2016 1:47 PM, "William Van Woensel" <William.Van.Woensel at dal.ca>
wrote:
>
> Hello Sebastian,
>
>
>
> I second Martynas' comment: the document is currently too abstract. Also
(and I mean this in the best possible way!), I would try to avoid using too
many buzzwords / name-dropping / fancy-sounding sentences without an
indication on their relevance to your work. It will greatly improve the
readability of the document.
>
I'll take note on this, thanks.
>
> If I'm not mistaken, your system accepts as input "plain" RDF (i.e.,
without any kind of ontological information), and, as a foundational
aspect, attempts to perform "type inferencing" to infer the missing
ontological information: initially, in particular, in the form of "kinds"
(i.e., including classes and meta-classes?), and later in the form of
contexts and behaviors. Perhaps it attempts to infer additional semantics
as well?
>
Yes, in the attempt of being able of using as many as possible datasources
/ formats I assume there is no schema information available.

Then I aggregate this inputs in three metamodel levels, one for
semiotically augmented triples and other for behavior (rules, flows and
events) inferred knowledge.
>
> A first question would be: why assume "plain" RDF? RDF is used exactly
for annotating resources with ontological knowledge. If you are considering
data sources of plain, unstructured text; works from the domain of ontology
extraction can be applied to (you guessed it) extract ontologies and thus
avoid having to deal with "plain" RDF. Similarly, plenty of work has been
done in extracting ontologies from relational databases. In both cases,
information / structure “hidden” in the original data source (text: nouns,
verbs, synonyms, etc.; rdb: columns, data types) are used to obtain an
ontology. So, one could say that your system works on a level of
abstraction that may be too high; it should go down and make sure that it
first obtains all available information from the original data source. If,
for some reason, you are stuck with "plain" RDF without type information,
you could have a look at formal concept analysis to extract concepts and
concept hierarchies (similar to what you mention in your document; this is
based on shared attribute values between objects).
>
FCA was one of my inspirations at the beginning. My conceptualization of
sets maybe could be seen as another way to represent concept lattices.

Inference (guess) of ontology / type information is one of my main attempts
/ goals. But as semantics is a branch of semiotics (C. S. Peirce) that
deals with signs and meanings I guessed there should be a 'sintactic' layer
(another semiotics branch) dealing with concepts, signs and objects and a
'pragmatics' layer (another semiotics branch) dealing with behavior and the
use of the previous constructions in the definition / declaration of, for
example, rules, events and state flows.
>
> Regarding “kinds”: it seems that your meta-classes could be modeled as
subtypes (Employee at XYZCorp > Employee) or via reification of the
“isEmployee” relation (isEmployee at XYZCorp). In that regard, it seems
like you’re trying to re-invent the wheel, but I could be wrong. So: what’s
the advantage of representing this information in your proposed way? The
Venn-diagrams in your documents are confusing, and it seems like you’re
indicating relations (e.g., subject > predicate-kind > object) instead of
overlaps between sets. But clearly, this aspect of “kinds” is another
foundational aspect of your work. Therefore, I would separate out this
aspect and start from scratch to 1) indicate what they precisely represent
(no wishy-washy statements, but rather concretely and formally define
them), and 2) explain the need for them, i.e., why they would be a useful
addition to meta-vocabularies such as RDF(S)/OWL.
>
The sets are overlapping.
"Kinds are the intersection of attributes and a values aggregated according
a given Resource triple occurrence. For the triple (a, b, c) if a is the
Resource (Subject), b and c are the attribute and value and all attributes
and values of the same Subject are aggregated with common attributes and
values from other resources to infer class and meta class relationships."

Kind classes are aggregated type (class) information. Kind metaclasses are
state (instances) information. Its not a subtype relation, is a kind of a
class relation.
>
> Anyway, just my two cents. I’ve seen your emails go past a number of
times now, and it seems that you’re going in circles (at least to me),
since each document looks pretty similar to the previous one. A complete
rewrite, focusing on one aspect at a time, could be of great benefit.
>

I'll be rewriting / refactoring the document soon. Thanks for your help,

Sebastián.

>
>
>
> Kind regards,
>
>
>
> William
>
>
>
>
>
>
>
> William Van Woensel PhD
>
> Post-Doctoral Fellow, Computer Science
>
> NICHE Research Group, Faculty of Computer Science
>
> Dalhousie University
> Halifax, NS B3H 4R2 Canada
> http://niche.cs.dal.ca/william-van-woensel
>
>
>
>
>
>
>
>
>
> -----Original Message-----
> From: Martynas Jusevičius [mailto:martynas at graphity.org]
> Sent: November-21-16 8:17 PM
> To: Sebastian Samaruga <ssamarug at gmail.com>
> Cc: pragmaticweb at lists.spline.inf.fu-berlin.de; Juan Sequeda <
juanfederico at gmail.com>; ProjectParadigm-ICT-Program <
metadataportals at yahoo.com>; Semantic Web <semantic-web at w3.org>; public-rww <
public-rww at w3.org>
> Subject: Re: Feedback
>
>
>
> Sebastian,
>
>
>
> please name actual datasources (Wikidata, UniProt, whatever),
vocabularies/ontologies (schema.org, Data Cube, etc.), data formats (XML,
CSV) that you want to use, and most importantly -- for what specific
purpose?
>
>
>
> Right now your document is so abstract it is incomprehensible and not
implementable.
>
>
>
> Martynas
>
>
>
> On Tue, Nov 22, 2016 at 1:06 AM, Sebastian Samaruga <ssamarug at gmail.com>
wrote:
>
> > Hi all, in response to Timothy's request I'll try to describe real
>
> > world problems / use cases I'm trying to solve: As the project I'd
>
> > like to be realized in this endeavor is a general purpose (knowledge
>
> > enabled) database back end with special features, use cases and
>
> > problems may be the same of the ones solved by traditional databases
>
> > but with semantic back end and special features provided benefits. So,
>
> > it will not do much by itself but to provide the means of higher
>
> > application / presentation layers taking advantage of such approaches.
>
> >
>
> > As the document I'm posting is kind of illegible stuff, I believe
>
> > sharing its link for comments will be of great help for me when
>
> > dumping my thoughts on the keyboard given useful advice is provided for
making things clearer.
>
> > Here is the Google Docs link (anyone can comment):
>
> >
>
> > https://docs.google.com/document/d/1mJbhTJSi907vrXfMtKly5biAMnoZJ5T-Kz
>
> > iaIMIELuM/edit?usp=drive_web
>
> >
>
> > Please be patient. I have this bunch of ideas, all low level, protocol
>
> > like (nothing like an 'application'), for back end and infrastructure
>
> > of concrete semantic applications. Maybe not even a little part of all
>
> > the document is worth reading material or is not well written. What
>
> > I'd like is finally get to communicate my concepts to see if it is
worth coding a 'proof of concept'
>
> > of this 'semantic services database'. The reason I'm so insistent in
>
> > having this feedback and potential consumers before I do some code is
>
> > that I've made so many attempts before by myself and I didn't get to
nothing alone.
>
> >
>
> > Best Regards,
>
> > Sebastián.
>
> >
>
> >
>
> > On Nov 19, 2016 7:58 PM, "Sebastian Samaruga" <ssamarug at gmail.com>
wrote:
>
> >>
>
> >> Hi,
>
> >>
>
> >> Trying to follow your advice I've added a Scope section at the
>
> >> beginning of the document. The reason why I've found so difficult to
>
> >> describe this 'application' is that it is not an application but it
>
> >> is more like a kind of
>
> >> (knowledge) backend database where (augmented) RDF and metamodels are
>
> >> my 'relational' model. I don't know if exists some kind of 'relational
algebra'
>
> >> for RDF so I started writing my own. Please tell me if I'm missing
>
> >> something important.
>
> >>
>
> >> Regards,
>
> >> Sebastián.
>
> >>
>
> >>
>
> >> On Nov 17, 2016 1:05 AM, "Juan Sequeda" <juanfederico at gmail.com> wrote:
>
> >>
>
> >> Sebastian,
>
> >>
>
> >> Writing advice I got early on:
>
> >>
>
> >> - First write an abstract. If you can't summarize in a few sentences
>
> >> what you are doing, then it is going to be very hard for other to
>
> >> understand
>
> >> - From the abstract, the following should be apparent
>
> >> 1) What is the problem
>
> >> 2) Why is it important (i.e. motivation)
>
> >> 3) What is your contribution (what is unique/novel)
>
> >>
>
> >> Your introduction should dive into a bit more detail on this.
>
> >>
>
> >> You should be answer each of these questions in a succinct and crisp
>
> >> sentence.
>
> >>
>
> >>
>
> >> --
>
> >> Juan Sequeda, Ph.D
>
> >> +1-575-SEQ-UEDA
>
> >> www.juansequeda.com
>
> >>
>
> >> On Mon, Nov 14, 2016 at 4:56 PM, Sebastian Samaruga
>
> >> <ssamarug at gmail.com>
>
> >> wrote:
>
> >>>
>
> >>> Hi all, its me again. I'm looking for feedback in this analysis
>
> >>> phase of a project I'd like to start building soon. The reason I
>
> >>> post this draft document again is that I've made some changes. I'd
>
> >>> like to have some orientation in the right directions I should take.
>
> >>> I hope not to be boring someone but 'cos what I'd like is to build
>
> >>> kind of augmented ontologies and metamodels, seems like no one is
willing to share this approach with me.
>
> >>>
>
> >>> Sorry if the document is a little rough written. I've wrote it all
>
> >>> on a phone...
>
> >>>
>
> >>> Best Regards,
>
> >>> Sebastián.
>
> >>
>
> >>
>
> >>
>
> >
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.spline.inf.fu-berlin.de/pipermail/pragmaticweb/attachments/20161122/afc146b6/attachment-0001.html>


More information about the PragmaticWeb mailing list